llama : support glm3 and glm4 (#8031 )

* add chatglm3-6b model support huggingface model: https://hf-mirror.com/THUDM/chatglm3-6b Signed-off-by: XingXing Qiao <qiaoxx@dingdao.com> * remove .rotary_pos_emb.inv_freq and unuse code for chatglm3 model Signed-off-by: XingXing Qiao <qiaoxx@dingdao.com> * fix lint error Signed-off-by: XingXing Qiao <qiaoxx@dingdao.com> * optimize convert-hf-to-gguf.py for chatglm model Signed-off-by: XingXing Qiao <qiaoxx@dingdao.com> * support glm-4-9b-chat Signed-off-by: XingXing Qiao <qiaoxx@dingdao.com> * fix eos tokens to glm4 * remove unused log * add preprocess to chatglm3 and chatglm4 * add eos_id_list to llama.cpp * fix code style * fix code style * fix conflicts * fix conflicts * Revert "add eos_id_list to llama.cpp" This reverts commit 3a4d5790bf. * set <|endoftext|> as eos and <|user|> as eot * fix chat template bug * add comment to glm prefix and suffix * fix conflicts and add rope_ratio & ChatGLMForConditionalGeneration * fix chat template bug * fix codestyle * fix conflicts * modified the general name of glm model * fix conflicts * remove prefix and suffix * use normal glm4 chattempalte & use LLM_FFN_SWIGLU in phi3 * fix: resolve Flake8 errors in `convert-hf-to-gguf.py` - Fix E302 by adding two blank lines before top-level function definitions - Replace print statements to fix NP100 - Fix E303 by ensuring only one blank line between lines of code * fix rope ratio to solve incorrect answers * fix by comments --------- Signed-off-by: XingXing Qiao <qiaoxx@dingdao.com> Co-authored-by: XingXing Qiao <qiaoxx@dingdao.com> Co-authored-by: Umpire2018 <138990495+Umpire2018@users.noreply.github.com>
llama : fix n_rot default (#8348 )
2026-07-01 01:57:43 +02:00 · 2024-07-07 15:52:10 +03:00 · 2024-07-07 14:59:02 +03:00 · 2024-07-07 14:23:38 +03:00 · 2024-07-07 13:38:02 +03:00 · 2024-07-07 13:37:47 +03:00
185 changed files with 8365 additions and 3686 deletions
@@ -27,7 +27,7 @@ COPY . .
 # Set nvcc architecture
 ENV CUDA_DOCKER_ARCH=${CUDA_DOCKER_ARCH}
 # Enable CUDA
-ENV LLAMA_CUDA=1
+ENV GGML_CUDA=1
 # Enable cURL
 ENV LLAMA_CURL=1

@@ -36,7 +36,7 @@ COPY . .
 # Set nvcc architecture
 ENV GPU_TARGETS=${ROCM_DOCKER_ARCH}
 # Enable ROCm
-ENV LLAMA_HIPBLAS=1
+ENV GGML_HIPBLAS=1
 ENV CC=/opt/rocm/llvm/bin/clang
 ENV CXX=/opt/rocm/llvm/bin/clang++

@@ -21,7 +21,7 @@ COPY . .
 # Set nvcc architecture
 ENV CUDA_DOCKER_ARCH=${CUDA_DOCKER_ARCH}
 # Enable CUDA
-ENV LLAMA_CUDA=1
+ENV GGML_CUDA=1

 RUN make -j$(nproc) llama-cli

@@ -2,7 +2,7 @@ ARG ONEAPI_VERSION=2024.1.1-devel-ubuntu22.04

 FROM intel/oneapi-basekit:$ONEAPI_VERSION as build

-ARG LLAMA_SYCL_F16=OFF
+ARG GGML_SYCL_F16=OFF
 RUN apt-get update && \
    apt-get install -y git

@@ -10,11 +10,11 @@ WORKDIR /app

 COPY . .

-RUN if [ "${LLAMA_SYCL_F16}" = "ON" ]; then \
-        echo "LLAMA_SYCL_F16 is set" && \
-        export OPT_SYCL_F16="-DLLAMA_SYCL_F16=ON"; \
+RUN if [ "${GGML_SYCL_F16}" = "ON" ]; then \
+        echo "GGML_SYCL_F16 is set" && \
+        export OPT_SYCL_F16="-DGGML_SYCL_F16=ON"; \
    fi && \
-    cmake -B build -DLLAMA_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx ${OPT_SYCL_F16} && \
+    cmake -B build -DGGML_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx ${OPT_SYCL_F16} && \
    cmake --build build --config Release --target llama-cli

 FROM intel/oneapi-basekit:$ONEAPI_VERSION as runtime
@@ -36,7 +36,7 @@ COPY . .
 # Set nvcc architecture
 ENV GPU_TARGETS=${ROCM_DOCKER_ARCH}
 # Enable ROCm
-ENV LLAMA_HIPBLAS=1
+ENV GGML_HIPBLAS=1
 ENV CC=/opt/rocm/llvm/bin/clang
 ENV CXX=/opt/rocm/llvm/bin/clang++

@@ -14,7 +14,7 @@ RUN wget -qO - https://packages.lunarg.com/lunarg-signing-key-pub.asc | apt-key
 # Build it
 WORKDIR /app
 COPY . .
-RUN cmake -B build -DLLAMA_VULKAN=1 && \
+RUN cmake -B build -DGGML_VULKAN=1 && \
    cmake --build build --config Release --target llama-cli

 # Clean up
@@ -1,84 +0,0 @@
-# SRPM for building from source and packaging an RPM for RPM-based distros.
-# https://docs.fedoraproject.org/en-US/quick-docs/creating-rpm-packages
-# Built and maintained by John Boero - boeroboy@gmail.com
-# In honor of Seth Vidal https://www.redhat.com/it/blog/thank-you-seth-vidal
-
-# Notes for llama.cpp:
-# 1. Tags are currently based on hash - which will not sort asciibetically.
-#    We need to declare standard versioning if people want to sort latest releases.
-# 2. Builds for CUDA/OpenCL support are separate, with different depenedencies.
-# 3. NVidia's developer repo must be enabled with nvcc, cublas, clblas, etc installed.
-#    Example: https://developer.download.nvidia.com/compute/cuda/repos/fedora37/x86_64/cuda-fedora37.repo
-# 4. OpenCL/CLBLAST support simply requires the ICD loader and basic opencl libraries.
-#    It is up to the user to install the correct vendor-specific support.
-
-Name:           llama.cpp-clblast
-Version:        %( date "+%%Y%%m%%d" )
-Release:        1%{?dist}
-Summary:        OpenCL Inference of LLaMA model in C/C++
-License:        MIT
-Source0:        https://github.com/ggerganov/llama.cpp/archive/refs/heads/master.tar.gz
-BuildRequires:  coreutils make gcc-c++ git mesa-libOpenCL-devel clblast-devel
-Requires:       clblast
-URL:            https://github.com/ggerganov/llama.cpp
-
-%define debug_package %{nil}
-%define source_date_epoch_from_changelog 0
-
-%description
-CPU inference for Meta's Lllama2 models using default options.
-
-%prep
-%setup -n llama.cpp-master
-
-%build
-make -j LLAMA_CLBLAST=1
-
-%install
-mkdir -p %{buildroot}%{_bindir}/
-cp -p llama-cli %{buildroot}%{_bindir}/llama-clblast-cli
-cp -p llama-server %{buildroot}%{_bindir}/llama-clblast-server
-cp -p llama-simple %{buildroot}%{_bindir}/llama-clblast-simple
-
-mkdir -p %{buildroot}/usr/lib/systemd/system
-%{__cat} <<EOF  > %{buildroot}/usr/lib/systemd/system/llamaclblast.service
-[Unit]
-Description=Llama.cpp server, CPU only (no GPU support in this build).
-After=syslog.target network.target local-fs.target remote-fs.target nss-lookup.target
-
-[Service]
-Type=simple
-EnvironmentFile=/etc/sysconfig/llama
-ExecStart=/usr/bin/llama-clblast-server $LLAMA_ARGS
-ExecReload=/bin/kill -s HUP $MAINPID
-Restart=never
-
-[Install]
-WantedBy=default.target
-EOF
-
-mkdir -p %{buildroot}/etc/sysconfig
-%{__cat} <<EOF  > %{buildroot}/etc/sysconfig/llama
-LLAMA_ARGS="-m /opt/llama2/ggml-model-f32.bin"
-EOF
-
-%clean
-rm -rf %{buildroot}
-rm -rf %{_builddir}/*
-
-%files
-%{_bindir}/llama-clblast-cli
-%{_bindir}/llama-clblast-server
-%{_bindir}/llama-clblast-simple
-/usr/lib/systemd/system/llamaclblast.service
-%config /etc/sysconfig/llama
-
-
-%pre
-
-%post
-
-%preun
-%postun
-
-%changelog
@@ -32,7 +32,7 @@ CPU inference for Meta's Lllama2 models using default options.
 %setup -n llama.cpp-master

 %build
-make -j LLAMA_CUDA=1
+make -j GGML_CUDA=1

 %install
 mkdir -p %{buildroot}%{_bindir}/
@@ -21,7 +21,7 @@ COPY . .
 # Set nvcc architecture
 ENV CUDA_DOCKER_ARCH=${CUDA_DOCKER_ARCH}
 # Enable CUDA
-ENV LLAMA_CUDA=1
+ENV GGML_CUDA=1
 # Enable cURL
 ENV LLAMA_CURL=1

@@ -2,7 +2,7 @@ ARG ONEAPI_VERSION=2024.1.1-devel-ubuntu22.04

 FROM intel/oneapi-basekit:$ONEAPI_VERSION as build

-ARG LLAMA_SYCL_F16=OFF
+ARG GGML_SYCL_F16=OFF
 RUN apt-get update && \
    apt-get install -y git libcurl4-openssl-dev

@@ -10,11 +10,11 @@ WORKDIR /app

 COPY . .

-RUN if [ "${LLAMA_SYCL_F16}" = "ON" ]; then \
-        echo "LLAMA_SYCL_F16 is set" && \
-        export OPT_SYCL_F16="-DLLAMA_SYCL_F16=ON"; \
+RUN if [ "${GGML_SYCL_F16}" = "ON" ]; then \
+        echo "GGML_SYCL_F16 is set" && \
+        export OPT_SYCL_F16="-DGGML_SYCL_F16=ON"; \
    fi && \
-    cmake -B build -DLLAMA_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx -DLLAMA_CURL=ON ${OPT_SYCL_F16} && \
+    cmake -B build -DGGML_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx -DLLAMA_CURL=ON ${OPT_SYCL_F16} && \
    cmake --build build --config Release --target llama-server

 FROM intel/oneapi-basekit:$ONEAPI_VERSION as runtime
@@ -36,7 +36,7 @@ COPY . .
 # Set nvcc architecture
 ENV GPU_TARGETS=${ROCM_DOCKER_ARCH}
 # Enable ROCm
-ENV LLAMA_HIPBLAS=1
+ENV GGML_HIPBLAS=1
 ENV CC=/opt/rocm/llvm/bin/clang
 ENV CXX=/opt/rocm/llvm/bin/clang++

@@ -14,7 +14,7 @@ RUN wget -qO - https://packages.lunarg.com/lunarg-signing-key-pub.asc | apt-key
 # Build it
 WORKDIR /app
 COPY . .
-RUN cmake -B build -DLLAMA_VULKAN=1 -DLLAMA_CURL=1 && \
+RUN cmake -B build -DGGML_VULKAN=1 -DLLAMA_CURL=1 && \
    cmake --build build --config Release --target llama-server

 # Clean up
@@ -17,19 +17,18 @@
  rocmPackages,
  vulkan-headers,
  vulkan-loader,
-  clblast,
+  curl,
  useBlas ? builtins.all (x: !x) [
    useCuda
    useMetalKit
-    useOpenCL
    useRocm
    useVulkan
  ] && blas.meta.available,
  useCuda ? config.cudaSupport,
-  useMetalKit ? stdenv.isAarch64 && stdenv.isDarwin && !useOpenCL,
+  useMetalKit ? stdenv.isAarch64 && stdenv.isDarwin,
  useMpi ? false, # Increases the runtime closure size by ~700M
-  useOpenCL ? false,
  useRocm ? config.rocmSupport,
+  enableCurl ? true,
  useVulkan ? false,
  llamaVersion ? "0.0.0", # Arbitrary version, substituted by the flake

@@ -56,7 +55,6 @@ let
    ++ lib.optionals useCuda [ "CUDA" ]
    ++ lib.optionals useMetalKit [ "MetalKit" ]
    ++ lib.optionals useMpi [ "MPI" ]
-    ++ lib.optionals useOpenCL [ "OpenCL" ]
    ++ lib.optionals useRocm [ "ROCm" ]
    ++ lib.optionals useVulkan [ "Vulkan" ];

@@ -198,19 +196,19 @@ effectiveStdenv.mkDerivation (
      optionals effectiveStdenv.isDarwin darwinBuildInputs
      ++ optionals useCuda cudaBuildInputs
      ++ optionals useMpi [ mpi ]
-      ++ optionals useOpenCL [ clblast ]
      ++ optionals useRocm rocmBuildInputs
      ++ optionals useBlas [ blas ]
-      ++ optionals useVulkan vulkanBuildInputs;
+      ++ optionals useVulkan vulkanBuildInputs
+      ++ optionals enableCurl [ curl ];

    cmakeFlags =
      [
        (cmakeBool "LLAMA_BUILD_SERVER" true)
        (cmakeBool "BUILD_SHARED_LIBS" (!enableStatic))
        (cmakeBool "CMAKE_SKIP_BUILD_RPATH" true)
+        (cmakeBool "LLAMA_CURL" enableCurl)
        (cmakeBool "GGML_NATIVE" false)
        (cmakeBool "GGML_BLAS" useBlas)
-        (cmakeBool "GGML_CLBLAST" useOpenCL)
        (cmakeBool "GGML_CUDA" useCuda)
        (cmakeBool "GGML_HIPBLAS" useRocm)
        (cmakeBool "GGML_METAL" useMetalKit)
@@ -254,7 +252,6 @@ effectiveStdenv.mkDerivation (
        useCuda
        useMetalKit
        useMpi
-        useOpenCL
        useRocm
        useVulkan
        ;
@@ -281,7 +278,7 @@ effectiveStdenv.mkDerivation (
      # Configurations we don't want even the CI to evaluate. Results in the
      # "unsupported platform" messages. This is mostly a no-op, because
      # cudaPackages would've refused to evaluate anyway.
-      badPlatforms = optionals (useCuda || useOpenCL) lib.platforms.darwin;
+      badPlatforms = optionals useCuda lib.platforms.darwin;

      # Configurations that are known to result in build failures. Can be
      # overridden by importing Nixpkgs with `allowBroken = true`.
@@ -9,5 +9,3 @@ contact_links:
  - name: Want to contribute?
    url: https://github.com/ggerganov/llama.cpp/wiki/contribute
    about: Head to the contribution guide page of the wiki for areas you can help with
-
-
@@ -10,10 +10,10 @@ on:
  push:
    branches:
      - master
-    paths: ['.github/workflows/**', '**/CMakeLists.txt', '**/Makefile', '**/*.h', '**/*.hpp', '**/*.c', '**/*.cpp', '**/*.cu', '**/*.swift', '**/*.m']
+    paths: ['.github/workflows/build.yml', '**/CMakeLists.txt', '**/Makefile', '**/*.h', '**/*.hpp', '**/*.c', '**/*.cpp', '**/*.cu', '**/*.cuh', '**/*.swift', '**/*.m', '**/*.metal']
  pull_request:
    types: [opened, synchronize, reopened]
-    paths: ['.github/workflows/build.yml', '**/CMakeLists.txt', '**/Makefile', '**/*.h', '**/*.hpp', '**/*.c', '**/*.cpp', '**/*.cu', '**/*.cuh', '**/*.swift', '**/*.m']
+    paths: ['.github/workflows/build.yml', '**/CMakeLists.txt', '**/Makefile', '**/*.h', '**/*.hpp', '**/*.c', '**/*.cpp', '**/*.cu', '**/*.cuh', '**/*.swift', '**/*.m', '**/*.metal']

 concurrency:
  group: ${{ github.workflow }}-${{ github.head_ref && github.ref || github.run_id }}
@@ -47,7 +47,7 @@ jobs:
          sysctl -a
          mkdir build
          cd build
-          cmake -DLLAMA_FATAL_WARNINGS=ON -DGGML_METAL_EMBED_LIBRARY=ON -DLLAMA_CURL=ON ..
+          cmake -DLLAMA_FATAL_WARNINGS=ON -DGGML_METAL_EMBED_LIBRARY=ON -DLLAMA_CURL=ON -DBUILD_SHARED_LIBS=OFF ..
          cmake --build . --config Release -j $(sysctl -n hw.logicalcpu)

      - name: Test
@@ -105,7 +105,7 @@ jobs:
          sysctl -a
          # Metal is disabled due to intermittent failures with Github runners not having a GPU:
          # https://github.com/ggerganov/llama.cpp/actions/runs/8635935781/job/23674807267#step:5:2313
-          cmake -B build -DLLAMA_FATAL_WARNINGS=ON -DGGML_METAL=OFF -DLLAMA_CURL=ON
+          cmake -B build -DLLAMA_FATAL_WARNINGS=ON -DGGML_METAL=OFF -DLLAMA_CURL=ON -DBUILD_SHARED_LIBS=OFF
          cmake --build build --config Release -j $(sysctl -n hw.logicalcpu)

      - name: Test
@@ -222,7 +222,7 @@ jobs:
        run: |
          mkdir build
          cd build
-          cmake .. -DLLAMA_FATAL_WARNINGS=ON -DLLAMA_CURL=ON
+          cmake .. -DLLAMA_FATAL_WARNINGS=ON -DLLAMA_CURL=ON -DBUILD_SHARED_LIBS=OFF
          cmake --build . --config Release -j $(nproc)

      - name: Test
@@ -799,6 +799,7 @@ jobs:
          7z x "-o${env:RUNNER_TEMP}" $env:RUNNER_TEMP/sde.tar
          $sde = $(join-path $env:RUNNER_TEMP sde-external-${env:SDE_VERSION}-win/sde.exe)
          cd build
+          $env:LLAMA_SKIP_TESTS_SLOW_ON_EMULATOR = 1
          & $sde -future -- ctest -L main -C Release --verbose --timeout 900

      - name: Determine tag name
@@ -14,6 +14,7 @@ on:
  push:
    branches:
      - master
+    paths: ['.github/workflows/docker.yml', '.devops/*.Dockerfile', '**/CMakeLists.txt', '**/Makefile', '**/*.h', '**/*.hpp', '**/*.c', '**/*.cpp', '**/*.cu', '**/*.cuh', '**/*.swift', '**/*.m', '**/*.metal']

 concurrency:
  group: ${{ github.workflow }}-${{ github.head_ref && github.ref || github.run_id }}
@@ -47,6 +47,7 @@ build*
 !build-info.cpp.in
 !build-info.sh
 !build.zig
+!docs/build.md
 /libllama.so
 /llama-*
 android-ndk-*
@@ -98,13 +99,14 @@ examples/server/*.mjs.hpp

 # Python

-__pycache__
-.venv
-/Pipfile
-dist
-poetry.lock
+/.venv
+__pycache__/
+*/poetry.lock
 poetry.toml

+# Nix
+/result
+
 # Test binaries
 /tests/test-backend-ops
 /tests/test-double-float
@@ -1,8 +1,9 @@
-# date: Tue Apr  9 09:17:14 EEST 2024
+# date: Wed Jun 26 19:36:34 EEST 2024
 # this file is auto-generated by scripts/gen-authors.sh

 0cc4m <picard12@live.de>
 0xspringtime <110655352+0xspringtime@users.noreply.github.com>
+20kdc <asdd2808@gmail.com>
 2f38b454 <dxf@protonmail.com>
 3ooabkhxtn <31479382+3ooabkhxtn@users.noreply.github.com>
 44670 <44670@users.noreply.github.com>
@@ -11,14 +12,18 @@ AT <manyoso@users.noreply.github.com>
 Aarni Koskela <akx@iki.fi>
 Aaron Miller <apage43@ninjawhale.com>
 Aaryaman Vasishta <aaryaman.vasishta@amd.com>
+Abheek Gulati <abheekg@hotmail.com>
 Abhilash Majumder <30946547+abhilash1910@users.noreply.github.com>
 Abhishek Gopinath K <31348521+overtunned@users.noreply.github.com>
 Adithya Balaji <adithya.b94@gmail.com>
 AdithyanI <adithyan.i4internet@gmail.com>
 Adrian <smith.adriane@gmail.com>
 Adrian Hesketh <a-h@users.noreply.github.com>
+Ahmet Zeer <ahmed.zeer@std.yildiz.edu.tr>
 AidanBeltonS <87009434+AidanBeltonS@users.noreply.github.com>
 Aisuko <urakiny@gmail.com>
+Akarshan Biswas <akarshanbiswas@fedoraproject.org>
+Albert Jin <albert.jin@gmail.com>
 Alberto <57916483+albbus-stack@users.noreply.github.com>
 Alex <awhill19@icloud.com>
 Alex Azarov <alex@azarov.by>
@@ -35,19 +40,24 @@ Ali Nehzat <ali.nehzat@thanks.dev>
 Ali Tariq <ali.tariq@10xengineers.ai>
 Alon <alonfaraj@gmail.com>
 AlpinDale <52078762+AlpinDale@users.noreply.github.com>
+Amir <amir_zia@outlook.com>
 AmirAli Mirian <37371367+amiralimi@users.noreply.github.com>
 Ananta Bastola <anantarajbastola@gmail.com>
 Anas Ahouzi <112881240+aahouzi@users.noreply.github.com>
 András Salamon <ott2@users.noreply.github.com>
 Andrei <abetlen@gmail.com>
 Andrew Canis <andrew.canis@gmail.com>
+Andrew Downing <andrew2085@gmail.com>
 Andrew Duffy <a10y@users.noreply.github.com>
 Andrew Godfrey <AndrewGodfrey@users.noreply.github.com>
+Andy Tai <andy-tai@users.noreply.github.com>
 Arik Poznanski <arikpoz@users.noreply.github.com>
 Artem <guinmoon@gmail.com>
+Artem Zinnatullin <ceo@abstractny.gay>
 Artyom Lebedev <vagran.ast@gmail.com>
 Asbjørn Olling <asbjornolling@gmail.com>
 Ásgeir Bjarni Ingvarsson <asgeir@fundinn.org>
+Ashish <1856117+ashishdatta@users.noreply.github.com>
 Ashok Gelal <401055+ashokgelal@users.noreply.github.com>
 Ashraful Islam <ashraful.meche@gmail.com>
 Atsushi Tatsuma <yoshoku@outlook.com>
@@ -57,35 +67,46 @@ BADR <contact@pythops.com>
 Bach Le <bach@bullno1.com>
 Bailey Chittle <39804642+bachittle@users.noreply.github.com>
 BarfingLemurs <128182951+BarfingLemurs@users.noreply.github.com>
+Bartowski <ckealty1182@gmail.com>
 Behnam M <58621210+ibehnam@users.noreply.github.com>
+Ben Ashbaugh <ben.ashbaugh@intel.com>
 Ben Garney <bengarney@users.noreply.github.com>
 Ben Siraphob <bensiraphob@gmail.com>
 Ben Williams <ben@719ben.com>
+Benjamin Findley <39356821+Kartoffelsaft@users.noreply.github.com>
 Benjamin Lecaillon <84293038+blecaillon@users.noreply.github.com>
 Bernat Vadell <hounter.caza@gmail.com>
+Bingan <70050083+binganao@users.noreply.github.com>
 Bodo Graumann <mail@bodograumann.de>
 Bono Lv <lvscar@users.noreply.github.com>
 Borislav Stanimirov <b.stanimirov@abv.bg>
 Branden Butler <bwtbutler@hotmail.com>
 Brian <mofosyne@gmail.com>
 Bruce MacDonald <brucewmacdonald@gmail.com>
+Bryan Honof <bryanhonof@gmail.com>
 CJ Pais <cj@cjpais.com>
 CRD716 <crd716@gmail.com>
+Calvin Laurenson <calvin@laurenson.dev>
 Cameron <csteele@steelecameron.com>
 Cameron Kaiser <classilla@users.noreply.github.com>
+Carolinabanana <140120812+Carolinabanana@users.noreply.github.com>
 Casey Primozic <casey@cprimozic.net>
 Casey Primozic <me@ameo.link>
 CausalLM <148736309+CausalLM@users.noreply.github.com>
 Cebtenzzre <cebtenzzre@gmail.com>
 Chad Brewbaker <crb002@gmail.com>
+Chao Jiang <jc19chaoj@zoho.com>
 Cheng Shao <terrorjack@type.dance>
+Chris Elrod <elrodc@gmail.com>
 Chris Kuehl <ckuehl@ckuehl.me>
 Christian Demsar <christian@github.email.demsar.us>
 Christian Demsar <crasm@git.vczf.us>
 Christian Falch <875252+chrfalch@users.noreply.github.com>
 Christian Kögler <ck3d@gmx.de>
+Christian Zhou-Zheng <59622928+christianazinn@users.noreply.github.com>
 Clark Saben <76020733+csaben@users.noreply.github.com>
 Clint Herron <hanclinto@gmail.com>
+CrispStrobe <154636388+CrispStrobe@users.noreply.github.com>
 Cuong Trinh Manh <nguoithichkhampha@gmail.com>
 DAN™ <dranger003@gmail.com>
 Damian Stewart <d@damianstewart.com>
@@ -95,8 +116,12 @@ Daniel Bevenius <daniel.bevenius@gmail.com>
 Daniel Drake <drake@endlessos.org>
 Daniel Hiltgen <dhiltgen@users.noreply.github.com>
 Daniel Illescas Romero <illescas.daniel@protonmail.com>
+Daniele <57776841+daniandtheweb@users.noreply.github.com>
 DannyDaemonic <DannyDaemonic@gmail.com>
 Dat Quoc Nguyen <2412555+datquocnguyen@users.noreply.github.com>
+Dave <dave-fl@users.noreply.github.com>
+Dave Airlie <airlied@gmail.com>
+Dave Airlie <airlied@redhat.com>
 Dave Della Costa <ddellacosta+github@gmail.com>
 David Friehs <david@friehs.info>
 David Kennedy <dakennedyd@gmail.com>
@@ -104,10 +129,13 @@ David Pflug <david@pflug.email>
 David Renshaw <dwrenshaw@gmail.com>
 David Sommers <12738+databyte@users.noreply.github.com>
 David Yang <davidyang6us@gmail.com>
+Dawid Potocki <github@dawidpotocki.com>
 Dawid Wysocki <62249621+TortillaZHawaii@users.noreply.github.com>
 Dean <Dean.Sinaean@gmail.com>
 Deins <deinsegle@gmail.com>
+Deven Mistry <31466137+deven367@users.noreply.github.com>
 Didzis Gosko <didzis@users.noreply.github.com>
+Djip007 <djip.perois@free.fr>
 Don Mahurin <dmahurin@users.noreply.github.com>
 DooWoong Lee (David) <manics99@naver.com>
 Doomsdayrs <38189170+Doomsdayrs@users.noreply.github.com>
@@ -116,8 +144,11 @@ Dr. Tom Murphy VII Ph.D <499244+tom7@users.noreply.github.com>
 Ebey Abraham <ebey97@gmail.com>
 Ed Lee <edilee@mozilla.com>
 Ed Lepedus <ed.lepedus@googlemail.com>
+Eddie-Wang <wangjinheng1120@163.com>
 Edward Taylor <edeetee@gmail.com>
+Elaine <elaine.zosa@gmail.com>
 Elbios <141279586+Elbios@users.noreply.github.com>
+Elton Kola <eltonkola@gmail.com>
 Engininja2 <139037756+Engininja2@users.noreply.github.com>
 Equim <sayaka@ekyu.moe>
 Eric Sommerlade <es0m@users.noreply.github.com>
@@ -143,37 +174,47 @@ Firat <firatkiral@gmail.com>
 Folko-Ven <71110216+Folko-Ven@users.noreply.github.com>
 Foul-Tarnished <107711110+Foul-Tarnished@users.noreply.github.com>
 Francisco Melo <43780565+francis2tm@users.noreply.github.com>
+Frank Mai <thxcode0824@gmail.com>
 FrankHB <frankhb1989@gmail.com>
+Fred Douglas <43351173+fredlas@users.noreply.github.com>
 Frederik Vogel <Schaltfehler@users.noreply.github.com>
 Gabe Goodhart <gabe.l.hart@gmail.com>
 GainLee <perfecter.gen@gmail.com>
 Galunid <karolek1231456@gmail.com>
 Gary Linscott <glinscott@gmail.com>
 Gary Mulder <gjmulder@gmail.com>
+Gavin Zhao <gavinzhaojw@protonmail.com>
 Genkagaku.GPT <hlhr202@163.com>
 Georgi Gerganov <ggerganov@gmail.com>
 Gilad S <giladgd@users.noreply.github.com>
+Giuseppe Scrivano <giuseppe@scrivano.org>
 GiviMAD <GiviMAD@users.noreply.github.com>
 Govlzkoy <gotope@users.noreply.github.com>
 Guillaume "Vermeille" Sanchez <Guillaume.V.Sanchez@gmail.com>
 Guillaume Wenzek <gwenzek@users.noreply.github.com>
 Guoteng <32697156+SolenoidWGT@users.noreply.github.com>
 Gustavo Rocha Dias <91472747+gustrd@users.noreply.github.com>
+Haggai Nuchi <h.nuchi@gmail.com>
 Halalaluyafail3 <55773281+Halalaluyafail3@users.noreply.github.com>
+Hamdoud Hakem <90524568+hamdoudhakem@users.noreply.github.com>
+HanishKVC <hanishkvc@gmail.com>
 Haohui Mai <ricetons@gmail.com>
 Haoxiang Fei <tonyfettes@tonyfettes.com>
 Harald Fernengel <harald.fernengel@here.com>
 Hatsune Miku <129688334+at8u@users.noreply.github.com>
+HatsuneMikuUwU33 <173229399+HatsuneMikuUwU33@users.noreply.github.com>
 Henk Poley <HenkPoley@gmail.com>
 Henri Vasserman <henv@hot.ee>
 Henrik Forstén <henrik.forsten@gmail.com>
 Herman Semenov <GermanAizek@yandex.ru>
 Hesen Peng <hesen.peng@gmail.com>
 Hoang Nguyen <hugo53@users.noreply.github.com>
+Hong Bo PENG <penghb@cn.ibm.com>
 Hongyu Ouyang <96765450+casavaca@users.noreply.github.com>
 Howard Su <howard0su@gmail.com>
 Hua Jiang <allenhjiang@outlook.com>
 Huawei Lin <huaweilin.cs@gmail.com>
+Hugo Roussel <hugo.rous@gmail.com>
 Ian Bull <irbull@eclipsesource.com>
 Ian Bull <irbull@gmail.com>
 Ian Scrivener <github@zilogy.asia>
@@ -190,8 +231,10 @@ Ivan Stepanov <ivanstepanovftw@gmail.com>
 JH23X <165871467+JH23X@users.noreply.github.com>
 Jack Mousseau <jmousseau@users.noreply.github.com>
 JackJollimore <130917767+JackJollimore@users.noreply.github.com>
+Jaemin Son <woalsdnd@gmail.com>
 Jag Chadha <jagtesh@gmail.com>
 Jakub N <jakubniemczyk97@gmail.com>
+James A Capozzoli <157492257+jac-jim@users.noreply.github.com>
 James Reynolds <magnusviri@users.noreply.github.com>
 Jan Boon <jan.boon@kaetemi.be>
 Jan Boon <kaetemi@gmail.com>
@@ -205,12 +248,17 @@ Jean-Michaël Celerier <jeanmichael.celerier+github@gmail.com>
 Jed Fox <git@jedfox.com>
 Jeffrey Quesnelle <emozilla@nousresearch.com>
 Jesse Jojo Johnson <williamsaintgeorge@gmail.com>
+Jeximo <jeximo@gmail.com>
 Jhen-Jie Hong <iainst0409@gmail.com>
 Jiahao Li <liplus17@163.com>
 Jian Liao <jianliao@users.noreply.github.com>
 JidongZhang-THU <1119708529@qq.com>
 Jinwoo Jeong <33892306+williamjeong2@users.noreply.github.com>
 Jiří Podivín <66251151+jpodivin@users.noreply.github.com>
+Jiří Sejkora <Sejseloid@gmail.com>
+Joan Fontanals <jfontanalsmartinez@gmail.com>
+Joan Fontanals <joan.fontanals.martinez@jina.ai>
+Johan <JohanAR@users.noreply.github.com>
 Johannes Gäßler <johannesg@5d6.de>
 Johannes Rudolph <johannes.rudolph@gmail.com>
 John <78893154+cmp-nct@users.noreply.github.com>
@@ -221,15 +269,19 @@ Jonas Wunderlich <32615971+jonas-w@users.noreply.github.com>
 Jorge A <161275481+jorgealias@users.noreply.github.com>
 Jose Maldonado <63384398+yukiteruamano@users.noreply.github.com>
 Joseph Stahl <1269177+josephst@users.noreply.github.com>
+Josh Ramer <josh.ramer@icloud.com>
 Joyce <joycebrum@google.com>
 Juan Calderon-Perez <835733+gaby@users.noreply.github.com>
 Judd <foldl@users.noreply.github.com>
 Julius Arkenberg <arki05@users.noreply.github.com>
 Jun Jie <71215065+junnjiee16@users.noreply.github.com>
+Junyang Lin <justinlin930319@hotmail.com>
 Juraj Bednar <juraj@bednar.io>
 Justin Parker <jparkerweb@gmail.com>
 Justin Suess <justin.suess@westpoint.edu>
+Justina Cho <justcho5@gmail.com>
 Justine Tunney <jtunney@gmail.com>
+Justine Tunney <jtunney@mozilla.com>
 Juuso Alasuutari <juuso.alasuutari@gmail.com>
 KASR <karim.asrih@gmail.com>
 Kamil Tomšík <info@tomsik.cz>
@@ -242,6 +294,7 @@ Kawrakow <48489457+ikawrakow@users.noreply.github.com>
 Keiichi Tabata <keiichi.tabata@outlook.com>
 Kenvix ⭐ <kenvixzure@live.com>
 Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
+Kevin Gibbons <bakkot@gmail.com>
 Kevin Ji <1146876+kevinji@users.noreply.github.com>
 Kevin Kwok <antimatter15@gmail.com>
 Kevin Lo <kevlo@kevlo.org>
@@ -257,6 +310,7 @@ Laura <Tijntje_7@msn.com>
 Lee <44310445+lx200916@users.noreply.github.com>
 Lee Drake <b.lee.drake@gmail.com>
 Leng Yue <lengyue@lengyue.me>
+Leon Knauer <git@leonknauer.com>
 LeonEricsson <70749762+LeonEricsson@users.noreply.github.com>
 Leonardo Neumann <leonardo@neumann.dev.br>
 Li Tan <tanliboy@gmail.com>
@@ -265,20 +319,26 @@ LoganDark <github@logandark.mozmail.com>
 LostRuins <39025047+LostRuins@users.noreply.github.com>
 Luciano <lucianostrika44@gmail.com>
 Luo Tian <lt@basecity.com>
+Lyle Dean <dean@lyle.dev>
 M. Yusuf Sarıgöz <yusufsarigoz@gmail.com>
 Maarten ter Huurne <maarten@treewalker.org>
 Mack Straight <eiz@users.noreply.github.com>
 Maël Kerbiriou <m431.kerbiriou@gmail.com>
 MaggotHATE <clay1326@gmail.com>
+Manuel <44313466+makuche@users.noreply.github.com>
 Marc Köhlbrugge <subscriptions@marckohlbrugge.com>
 Marco Matthies <71844+marcom@users.noreply.github.com>
 Marcus Dunn <51931484+MarcusDunn@users.noreply.github.com>
 Marian Cepok <marian.cepok@gmail.com>
 Mark Fairbairn <thebaron88@gmail.com>
 Marko Tasic <mtasic85@gmail.com>
+Markus Tavenrath <mtavenrath@users.noreply.github.com>
+Martin Delille <martin@delille.org>
 Martin Krasser <krasserm@googlemail.com>
 Martin Schwaighofer <mschwaig@users.noreply.github.com>
 Marvin Gießing <marvin.giessing@gmail.com>
+Masaya, Kato <62578291+msy-kato@users.noreply.github.com>
+MasterYi1024 <39848311+MasterYi1024@users.noreply.github.com>
 Mateusz Charytoniuk <mateusz.charytoniuk@protonmail.com>
 Matheus C. França <matheus-catarino@hotmail.com>
 Matheus Gabriel Alves Silva <matheusgasource@gmail.com>
@@ -287,8 +347,11 @@ Mathijs de Bruin <mathijs@mathijsfietst.nl>
 Matt Clayton <156335168+mattjcly@users.noreply.github.com>
 Matt Pulver <matt.pulver@heavy.ai>
 Matteo Boschini <12133566+mbosc@users.noreply.github.com>
+Mattheus Chediak <shammcity00@gmail.com>
 Matthew Tejo <matthew.tejo@gmail.com>
 Matvey Soloviev <blackhole89@gmail.com>
+Max Krasnyansky <max.krasnyansky@gmail.com>
+Max Krasnyansky <quic_maxk@quicinc.com>
 Maxime <672982+maximegmd@users.noreply.github.com>
 Maximilian Winter <maximilian.winter.91@gmail.com>
 Meng Zhang <meng@tabbyml.com>
@@ -300,32 +363,41 @@ Michael Kesper <mkesper@schokokeks.org>
 Michael Klimenko <mklimenko29@gmail.com>
 Michael Podvitskiy <podvitskiymichael@gmail.com>
 Michael Potter <NanoTekGuy@Gmail.com>
+Michael de Gans <michael.john.degans@gmail.com>
 Michaël de Vries <vriesdemichael@gmail.com>
 Mihai <mihai.chirculescu@yahoo.com>
 Mike <ytianhui2004@gmail.com>
+Mikko Juola <mikjuo@gmail.com>
 Minsoo Cheong <54794500+mscheong01@users.noreply.github.com>
 Mirko185 <mirkosig@gmail.com>
 Mirror Azure <54669636+MirrorAzure@users.noreply.github.com>
 Miwa / Ensan <63481257+ensan-hcl@users.noreply.github.com>
 Mohammadreza Hendiani <hendiani.mohammadreza@gmail.com>
+Mohammadreza Hendiani <mohammad.r.hendiani@gmail.com>
 Murilo Santana <mvrilo@gmail.com>
 Musab Gultekin <musabgultekin@users.noreply.github.com>
 Nam D. Tran <42194884+namtranase@users.noreply.github.com>
+Nathan Epstein <nate2@umbc.edu>
 NawafAlansari <72708095+NawafAlansari@users.noreply.github.com>
 Nebula <infinitewormhole@gmail.com>
+Neo Zhang <14088817+arthw@users.noreply.github.com>
+Neo Zhang <zhang.jianyu@outlook.com>
 Neo Zhang Jianyu <jianyu.zhang@intel.com>
 Neuman Vong <neuman.vong@gmail.com>
 Nexesenex <124105151+Nexesenex@users.noreply.github.com>
 Niall Coates <1349685+Niall-@users.noreply.github.com>
 Nicolai Weitkemper <kontakt@nicolaiweitkemper.de>
+Nicolás Pérez <nicolas_perez@brown.edu>
 Nigel Bosch <pnigelb@gmail.com>
 Niklas Korz <niklas@niklaskorz.de>
+Nikolas <127742645+nneubacher@users.noreply.github.com>
 Nindaleth <Nindaleth@users.noreply.github.com>
 Oleksandr Nikitin <oleksandr@tvori.info>
 Oleksii Maryshchenko <oleksii.maryshchenko@gmail.com>
 Olivier Chafik <ochafik@users.noreply.github.com>
 Ondřej Čertík <ondrej@certik.us>
 Ouadie EL FAROUKI <ouadie.elfarouki@codeplay.com>
+Patrice Ferlet <metal3d@gmail.com>
 Paul Tsochantaris <ptsochantaris@icloud.com>
 Pavol Rusnak <pavol@rusnak.io>
 Pedro Cuenca <pedro@huggingface.co>
@@ -343,9 +415,14 @@ RJ Adriaansen <adriaansen@eshcc.eur.nl>
 Radoslav Gerganov <rgerganov@gmail.com>
 Radosław Gryta <radek.gryta@gmail.com>
 Rahul Vivek Nair <68507071+RahulVivekNair@users.noreply.github.com>
+Raj Hammeer Singh Hada <hammeerraj@gmail.com>
+Ralph Soika <ralph.soika@imixs.com>
 Rand Xie <randxiexyy29@gmail.com>
 Randall Fitzgerald <randall@dasaku.net>
 Reinforce-II <fate@eastal.com>
+Ren Xuancheng <jklj077@users.noreply.github.com>
+Rene Leonhardt <65483435+reneleonhardt@users.noreply.github.com>
+RhinoDevel <RhinoDevel@users.noreply.github.com>
 Riceball LEE <snowyu.lee@gmail.com>
 Richard Kiss <him@richardkiss.com>
 Richard Roberson <richardr1126@gmail.com>
@@ -373,6 +450,7 @@ Rowan Hart <rowanbhart@gmail.com>
 Rune <43761327+Rune-AI@users.noreply.github.com>
 Ryan Landay <rlanday@gmail.com>
 Ryder Wishart <ryderwishart@gmail.com>
+Ryuei <louixs@users.noreply.github.com>
 Rőczey Barnabás <31726601+An0nie@users.noreply.github.com>
 SakuraUmi <yukinon244@gmail.com>
 Salvador E. Tropea <stropea@inti.gob.ar>
@@ -386,6 +464,7 @@ SebastianApel <13675545+SebastianApel@users.noreply.github.com>
 Senemu <10880819+Senemu@users.noreply.github.com>
 Sergey Alirzaev <zl29ah@gmail.com>
 Sergio López <slp@sinrega.org>
+Sertaç Özercan <852750+sozercan@users.noreply.github.com>
 SeungWon Jeong <65549245+redlion0929@users.noreply.github.com>
 ShadovvBeast <ShadovvBeast@gmail.com>
 Shakhar Dasgupta <shakhardasgupta@gmail.com>
@@ -394,6 +473,7 @@ Shijie <821898965@qq.com>
 Shintarou Okada <kokuzen@gmail.com>
 Shouzheng Liu <61452103+lshzh-ww@users.noreply.github.com>
 Shouzheng Liu <lshzh.hi@gmail.com>
+Shuichi Tsutsumi <shuichi0526@gmail.com>
 Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>
 Simon Willison <swillison@gmail.com>
 Siwen Yu <yusiwen@gmail.com>
@@ -405,11 +485,14 @@ Someone <sergei.kozlukov@aalto.fi>
 Someone Serge <sergei.kozlukov@aalto.fi>
 Sourab Mangrulkar <13534540+pacman100@users.noreply.github.com>
 Spencer Sutton <spencersutton@users.noreply.github.com>
+Srihari-mcw <96763064+Srihari-mcw@users.noreply.github.com>
 Srinivas Billa <nivibilla@gmail.com>
 Stefan Sydow <stefan@sydow.email>
+Steffen Röcker <sroecker@gmail.com>
 Stephan Walter <stephan@walter.name>
 Stephen Nichols <snichols@users.noreply.github.com>
 Steve Grubb <ausearch.1@gmail.com>
+Steven Prichard <spprichard20@gmail.com>
 Steven Roussey <sroussey@gmail.com>
 Steward Garcia <57494570+FSSRepo@users.noreply.github.com>
 Suaj Carrot <72162667+SuajCarrot@users.noreply.github.com>
@@ -434,16 +517,19 @@ Tom C <tom.corelis@gmail.com>
 Tom Jobbins <784313+TheBloke@users.noreply.github.com>
 Tomas <tom.tomas.36478119@gmail.com>
 Tomáš Pazdiora <tomas.pazdiora@gmail.com>
+Tristan Druyen <tristan@vault81.mozmail.com>
 Tristan Ross <rosscomputerguy@protonmail.com>
 Tungsten842 <886724vf@anonaddy.me>
 Tungsten842 <quantmint@protonmail.com>
 Tushar <ditsuke@protonmail.com>
 UEXTM.com <84163508+uextm@users.noreply.github.com>
+Ulrich Drepper <drepper@gmail.com>
 Uzo Nweke <uzoechi@gmail.com>
 Vaibhav Srivastav <vaibhavs10@gmail.com>
 Val Kharitonov <mail@kharvd.com>
 Valentin Konovalov <valle.ketsujin@gmail.com>
 Valentyn Bezshapkin <61702053+valentynbez@users.noreply.github.com>
+Victor Nogueira <felladrin@gmail.com>
 Victor Z. Peng <ziliangdotme@gmail.com>
 Vlad <spitfireage@gmail.com>
 Vladimir <bogdad@gmail.com>
@@ -455,7 +541,9 @@ Weird Constructor <weirdconstructor@gmail.com>
 Welby Seely <welbyseely@gmail.com>
 Wentai Zhang <rchardx@gmail.com>
 WillCorticesAI <150854901+WillCorticesAI@users.noreply.github.com>
+William Tambellini <william.tambellini@gmail.com>
 Willy Tarreau <w@1wt.eu>
+Wouter <9594229+DifferentialityDevelopment@users.noreply.github.com>
 Wu Jian Ping <wujjpp@hotmail.com>
 Wu Jian Ping <wujp@greatld.com>
 Xiake Sun <xiake.sun@intel.com>
@@ -466,6 +554,8 @@ Xiaoyi Chen <cxychina@gmail.com>
 Xingchen Song(宋星辰) <xingchensong1996@163.com>
 Xuan Son Nguyen <thichthat@gmail.com>
 Yann Follet <131855179+YannFollet@users.noreply.github.com>
+Yaroslav <yaroslav.yashin@me.com>
+Yazan Agha-Schrader <mountaiin@icloud.com>
 Yiming Cui <conandiy@vip.qq.com>
 Yishuo Wang <MeouSker77@outlook.com>
 Yueh-Po Peng <94939112+y10ab1@users.noreply.github.com>
@@ -477,6 +567,7 @@ Zane Shannon <z@zcs.me>
 Zay <95888118+isaiahbjork@users.noreply.github.com>
 Zenix <zenixls2@gmail.com>
 Zhang Peiyuan <a1286225768@gmail.com>
+Zheng.Deng <32841220+dengzheng-cloud@users.noreply.github.com>
 ZhouYuChen <zhouyuchen@naver.com>
 Ziad Ben Hadj-Alouane <zied.benhadjalouane@gmail.com>
 Ziang Wu <97337387+ZiangWu-77@users.noreply.github.com>
@@ -484,14 +575,18 @@ Zsapi <martin1.zsapka@gmail.com>
 a-n-n-a-l-e-e <150648636+a-n-n-a-l-e-e@users.noreply.github.com>
 adel boussaken <netdur@gmail.com>
 afrideva <95653597+afrideva@users.noreply.github.com>
+agray3 <agray3@users.noreply.github.com>
 akawrykow <142945436+akawrykow@users.noreply.github.com>
 alexpinel <93524949+alexpinel@users.noreply.github.com>
 alonfaraj <alonfaraj@gmail.com>
+alwqx <kenan3015@gmail.com>
+amd-lalithnc <lalithnc@amd.com>
 andrijdavid <david@geek.mg>
 anon998 <131767832+anon998@users.noreply.github.com>
 anzz1 <anzz1@live.com>
 apaz <aarpazdera@gmail.com>
 apcameron <37645737+apcameron@users.noreply.github.com>
+arch-btw <57669023+arch-btw@users.noreply.github.com>
 arcrank <arcrank@gmail.com>
 arlo-phoenix <140345165+arlo-phoenix@users.noreply.github.com>
 at8u <129688334+at8u@users.noreply.github.com>
@@ -514,13 +609,17 @@ cocktailpeanut <121128867+cocktailpeanut@users.noreply.github.com>
 coezbek <c.oezbek@gmail.com>
 comex <comexk@gmail.com>
 compilade <113953597+compilade@users.noreply.github.com>
+compilade <git@compilade.net>
+cpumaxx <163466046+cpumaxx@users.noreply.github.com>
 crasm <crasm@git.vczf.net>
 crasm <crasm@git.vczf.us>
 daboe01 <daboe01@googlemail.com>
 david raistrick <keen99@users.noreply.github.com>
+ddh0 <dylanhalladay02@icloud.com>
 ddpasa <112642920+ddpasa@users.noreply.github.com>
 deepdiffuser <112834445+deepdiffuser@users.noreply.github.com>
 divinity76 <divinity76@gmail.com>
+dm4 <sunrisedm4@gmail.com>
 dotpy314 <33351922+dotpy314@users.noreply.github.com>
 drbh <david.richard.holtz@gmail.com>
 ds5t5 <145942675+ds5t5@users.noreply.github.com>
@@ -529,6 +628,7 @@ eastriver <lee@eastriver.dev>
 ebraminio <ebraminio@gmail.com>
 eiery <19350831+eiery@users.noreply.github.com>
 eric8607242 <e0928021388@gmail.com>
+fairydreaming <166155368+fairydreaming@users.noreply.github.com>
 fraxy-v <65565042+fraxy-v@users.noreply.github.com>
 github-actions[bot] <github-actions[bot]@users.noreply.github.com>
 gliptic <gliptic@users.noreply.github.com>
@@ -539,6 +639,7 @@ h-h-h-h <13482553+h-h-h-h@users.noreply.github.com>
 hankcs <cnhankmc@gmail.com>
 hoangmit <hoangmit@users.noreply.github.com>
 hongbo.mo <352280764@qq.com>
+hopkins385 <98618192+hopkins385@users.noreply.github.com>
 howlger <eclipse@voormann.de>
 howlger <github@voormann.de>
 hutli <6594598+hutli@users.noreply.github.com>
@@ -549,14 +650,22 @@ hydai <z54981220@gmail.com>
 iSma <ismail.senhaji@gmail.com>
 iacore <74560659+iacore@users.noreply.github.com>
 igarnier <igarnier@protonmail.com>
+intelmatt <61025942+intelmatt@users.noreply.github.com>
 iohub <rickyang.pro@gmail.com>
 jacobi petrucciani <8117202+jpetrucciani@users.noreply.github.com>
+jaime-m-p <167997752+jaime-m-p@users.noreply.github.com>
 jameswu2014 <545426914@qq.com>
+jiez <373447296@qq.com>
 jneem <joeneeman@gmail.com>
+joecryptotoo <80373433+joecryptotoo@users.noreply.github.com>
 johnson442 <56517414+johnson442@users.noreply.github.com>
+jojorne <jojorne@users.noreply.github.com>
 jon-chuang <9093549+jon-chuang@users.noreply.github.com>
 jp-x-g <jpxg-dev@protonmail.com>
+jukofyork <69222624+jukofyork@users.noreply.github.com>
+junchao-loongson <68935141+junchao-loongson@users.noreply.github.com>
 jwj7140 <32943891+jwj7140@users.noreply.github.com>
+k.h.lai <adrian.k.h.lai@outlook.com>
 kaizau <kaizau@users.noreply.github.com>
 kalomaze <66376113+kalomaze@users.noreply.github.com>
 kang <tpdns9032100@gmail.com>
@@ -575,11 +684,15 @@ ldwang <ftgreat@163.com>
 le.chang <cljs118@126.com>
 leejet <leejet714@gmail.com>
 limitedAtonement <limitedAtonement@users.noreply.github.com>
+liuwei-git <14815172+liuwei-git@users.noreply.github.com>
 lon <114724657+longregen@users.noreply.github.com>
+loonerin <132926317+loonerin@users.noreply.github.com>
+luoyu-intel <yu.luo@intel.com>
 m3ndax <adrian.goessl@outlook.com>
 maddes8cht <55592906+maddes8cht@users.noreply.github.com>
 makomk <makosoft@googlemail.com>
 manikbhandari <mbbhandarimanik2@gmail.com>
+maor-ps <154728172+maor-ps@users.noreply.github.com>
 mdrokz <mohammadmunshi@gmail.com>
 mgroeber9110 <45620825+mgroeber9110@users.noreply.github.com>
 minarchist <minarchist@users.noreply.github.com>
@@ -593,15 +706,19 @@ ngc92 <7938269+ngc92@users.noreply.github.com>
 nhamanasu <45545786+nhamanasu@users.noreply.github.com>
 niansa/tuxifan <anton-sa@web.de>
 niansa/tuxifan <tuxifan@posteo.de>
+nickp27 <nb.porter@gmail.com>
 ningshanwutuobang <ningshanwutuobang@gmail.com>
 nold <Nold360@users.noreply.github.com>
 nopperl <54780682+nopperl@users.noreply.github.com>
 nusu-github <29514220+nusu-github@users.noreply.github.com>
 olexiyb <olexiyb@gmail.com>
+omahs <73983677+omahs@users.noreply.github.com>
 oobabooga <112222186+oobabooga@users.noreply.github.com>
 opparco <parco.opaai@gmail.com>
 ostix360 <55257054+ostix360@users.noreply.github.com>
+pengxin99 <pengxin.yuan@intel.com>
 perserk <perserk@gmail.com>
+pmysl <piotr.myslinski@outlook.com>
 postmasters <namnguyen@google.com>
 pudepiedj <pudepiedj@gmail.com>
 qingfengfenga <41416092+qingfengfenga@users.noreply.github.com>
@@ -614,16 +731,19 @@ rhuddleston <ryan.huddleston@percona.com>
 rimoliga <53384203+rimoliga@users.noreply.github.com>
 runfuture <runfuture@users.noreply.github.com>
 sandyiscool <sandyiscool@gmail.com>
+sasha0552 <admin@sasha0552.org>
 semidark <me@semidark.net>
 sharpHL <132747147+sharpHL@users.noreply.github.com>
 shibe2 <shibe@tuta.io>
 singularity <12184989+singularity-s0@users.noreply.github.com>
 sjinzh <sjinzh@gmail.com>
+sjxx <63994076+ylsdamxssjxxdd@users.noreply.github.com>
 slaren <2141330+slaren@users.noreply.github.com>
 slaren <slarengh@gmail.com>
 snadampal <87143774+snadampal@users.noreply.github.com>
 staviq <staviq@gmail.com>
 stduhpf <stephduh@live.fr>
+strawberrymelonpanda <152940198+strawberrymelonpanda@users.noreply.github.com>
 swittk <switt1995@gmail.com>
 takov751 <40316768+takov751@users.noreply.github.com>
 tarcey <cey.tarik@gmail.com>
@@ -636,12 +756,16 @@ uint256_t <konndennsa@gmail.com>
 uint256_t <maekawatoshiki1017@gmail.com>
 unbounded <haakon@likedan.net>
 valiray <133289098+valiray@users.noreply.github.com>
+vik <vikhyatk@gmail.com>
+viric <viric@viric.name>
 vodkaslime <646329483@qq.com>
 vvhg1 <94630311+vvhg1@users.noreply.github.com>
 vxiiduu <73044267+vxiiduu@users.noreply.github.com>
 wbpxre150 <100937007+wbpxre150@users.noreply.github.com>
 whoreson <139810751+whoreson@users.noreply.github.com>
+woachk <24752637+woachk@users.noreply.github.com>
 wonjun Jang <strutive07@gmail.com>
+woodx <124784234+woodx9@users.noreply.github.com>
 wzy <32936898+Freed-Wu@users.noreply.github.com>
 xaedes <xaedes@gmail.com>
 xaedes <xaedes@googlemail.com>
@@ -649,7 +773,10 @@ xloem <0xloem@gmail.com>
 yangli2 <yangli2@gmail.com>
 yuiseki <yuiseki@gmail.com>
 zakkor <edward.partenie@gmail.com>
+zhangkaihuo <zhangkaihuo@gmail.com>
 zhouwg <6889919+zhouwg@users.noreply.github.com>
+zhouwg <zhouwg2000@gmail.com>
 zrm <trustiosity.zrm@gmail.com>
+Ștefan-Gabriel Muscalu <legraphista@users.noreply.github.com>
 源文雨 <41315874+fumiama@users.noreply.github.com>
 Нияз Гарифзянов <112617865+garrnizon@users.noreply.github.com>
@@ -42,6 +42,10 @@ endif()

 option(BUILD_SHARED_LIBS "build shared libraries" ${BUILD_SHARED_LIBS_DEFAULT})

+if (WIN32)
+    add_compile_definitions(_CRT_SECURE_NO_WARNINGS)
+endif()
+
 #
 # option list
 #
@@ -79,13 +83,21 @@ set(GGML_SANITIZE_ADDRESS   ${LLAMA_SANITIZE_ADDRESS})
 set(GGML_SANITIZE_UNDEFINED ${LLAMA_SANITIZE_UNDEFINED})
 set(GGML_ALL_WARNINGS       ${LLAMA_ALL_WARNINGS})
 set(GGML_FATAL_WARNINGS     ${LLAMA_FATAL_WARNINGS})
-set(GGML_LLAMAFILE          ON)
+
+# change the default for these ggml options
+if (NOT DEFINED GGML_LLAMAFILE)
+    set(GGML_LLAMAFILE ON)
+endif()
+
+if (NOT DEFINED GGML_CUDA_USE_GRAPHS)
+    set(GGML_CUDA_USE_GRAPHS ON)
+endif()

 # transition helpers
 function (llama_option_depr TYPE OLD NEW)
    if (${OLD})
        message(${TYPE} "${OLD} is deprecated and will be removed in the future.\nUse ${NEW} instead\n")
-        set(${NEW} ON)
+        set(${NEW} ON PARENT_SCOPE)
    endif()
 endfunction()

@@ -95,7 +107,6 @@ llama_option_depr(WARNING     LLAMA_KOMPUTE             GGML_KOMPUTE)
 llama_option_depr(WARNING     LLAMA_METAL               GGML_METAL)
 llama_option_depr(WARNING     LLAMA_METAL_EMBED_LIBRARY GGML_METAL_EMBED_LIBRARY)
 llama_option_depr(WARNING     LLAMA_NATIVE              GGML_NATIVE)
-llama_option_depr(WARNING     LLAMA_OPENMP              GGML_OPENMP)
 llama_option_depr(WARNING     LLAMA_RPC                 GGML_RPC)
 llama_option_depr(WARNING     LLAMA_SYCL                GGML_SYCL)
 llama_option_depr(WARNING     LLAMA_SYCL_F16            GGML_SYCL_F16)
@@ -145,7 +156,7 @@ install(FILES ${CMAKE_CURRENT_BINARY_DIR}/llama-config.cmake
        DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/llama)

 install(
-    FILES convert-hf-to-gguf.py
+    FILES convert_hf_to_gguf.py
    PERMISSIONS
        OWNER_READ
        OWNER_WRITE
@@ -19,6 +19,7 @@
        "cacheVariables": {
            "CMAKE_EXPORT_COMPILE_COMMANDS": "ON",
            "CMAKE_CXX_COMPILER": "icx",
+            "CMAKE_C_COMPILER": "cl",
            "GGML_SYCL": "ON",
            "CMAKE_INSTALL_RPATH": "$ORIGIN;$ORIGIN/.."
        }
@@ -1,14 +1,24 @@
-# Contributing Guidelines
+# Pull requests

-## Checklist
+- Always squash-merge the PR before merging
+- Use the following format for your final commit: `<module> : <commit title> (#<issue_number>)`. For example: `utils : fix typo in utils.py (#1234)`
+- Test your changes:
+  - Using the commands in the [`tests`](tests) folder. For instance, running the `./tests/test-backend-ops` command tests different backend implementations of the GGML library
+  - Execute [the full CI locally on your machine](ci/README.md) before publishing
+- If the pull request contains only documentation changes (e.g., updating READMEs, adding new wiki pages), please add `[no ci]` to the commit title. This will skip unnecessary CI checks and help reduce build times
+- Please rate the complexity of your PR (i.e. `Review Complexity : Low`, `Review Complexity : Medium`, `Review Complexity : High`). This makes it easier for maintainers to triage the PRs.
+  - The PR template has a series of review complexity checkboxes `[ ]` that [you can mark as](https://docs.github.com/en/get-started/writing-on-github/working-with-advanced-formatting/about-task-lists) `[X]` for your conveience

-* Make sure your PR follows the [coding guidelines](https://github.com/ggerganov/llama.cpp/blob/master/README.md#coding-guidelines)
-* Test your changes using the commands in the [`tests`](tests) folder. For instance, running the `./tests/test-backend-ops` command tests different backend implementations of the GGML library
-* Execute [the full CI locally on your machine](ci/README.md) before publishing
+# Coding guidelines

-## PR formatting
+- Avoid adding third-party dependencies, extra files, extra headers, etc.
+- Always consider cross-compatibility with other operating systems and architectures
+- Avoid fancy looking modern STL constructs, use basic `for` loops, avoid templates, keep it simple
+- There are no strict rules for the code style, but try to follow the patterns in the code (indentation, spaces, etc.). Vertical alignment makes things more readable and easier to batch edit
+- Clean-up any trailing whitespaces, use 4 spaces for indentation, brackets on the same line, `void * ptr`, `int & a`
+- Naming usually optimizes for common prefix (see https://github.com/ggerganov/ggml/pull/302#discussion_r1243240963)
+- Tensors store data in row-major order. We refer to dimension 0 as columns, 1 as rows, 2 as matrices
+- Matrix multiplication is unconventional: [`C = ggml_mul_mat(ctx, A, B)`](https://github.com/ggerganov/llama.cpp/blob/880e352277fc017df4d5794f0c21c44e1eae2b84/ggml.h#L1058-L1064) means $C^T = A B^T \Leftrightarrow C = B A^T.$
+
+![matmul](media/matmul.png)

-* Please rate the complexity of your PR (i.e. `Review Complexity : Low`, `Review Complexity : Medium`, `Review Complexity : High`). This makes it easier for maintainers to triage the PRs.
-    - The PR template has a series of review complexity checkboxes `[ ]` that you can mark as `[X]` for your conveience. Refer to [About task lists](https://docs.github.com/en/get-started/writing-on-github/working-with-advanced-formatting/about-task-lists) for more information.
-* If the pull request only contains documentation changes (e.g., updating READMEs, adding new wiki pages), please add `[no ci]` to the commit title. This will skip unnecessary CI checks and help reduce build times.
-* When squashing multiple commits on merge, use the following format for your commit title: `<module> : <commit title> (#<issue_number>)`. For example: `utils : Fix typo in utils.py (#1234)`
@@ -45,6 +45,7 @@ BUILD_TARGETS = \
 TEST_TARGETS = \
 	tests/test-autorelease \
 	tests/test-backend-ops \
+	tests/test-chat-template \
 	tests/test-double-float \
 	tests/test-grad0 \
 	tests/test-grammar-integration \
@@ -61,6 +62,11 @@ TEST_TARGETS = \
 	tests/test-tokenizer-1-bpe \
 	tests/test-tokenizer-1-spm

+# Legacy build targets that were renamed in #7809, but should still be removed when the project is cleaned
+LEGACY_TARGETS = main quantize quantize-stats perplexity imatrix embedding vdot q8dot train-text-from-scratch convert-llama2c-to-ggml \
+	simple batched batched-bench save-load-state server gguf gguf-split eval-callback llama-bench libllava.a llava-cli baby-llama \
+	retrieval speculative infill tokenize benchmark-matmult parallel finetune export-lora lookahead lookup passkey gritlm
+
 # Deprecation aliases
 ifdef LLAMA_CUBLAS
 $(error LLAMA_CUBLAS is removed. Use GGML_CUDA instead.)
@@ -148,12 +154,6 @@ ifndef UNAME_M
 UNAME_M := $(shell uname -m)
 endif

-MK_CFLAGS     += -O3
-MK_CXXFLAGS   += -O3
-ifndef LLAMA_DEBUG
-MK_NVCCFLAGS  += -O3
-endif # LLAMA_DEBUG
-
 # In GNU make default CXX is g++ instead of c++.  Let's fix that so that users
 # of non-gcc compilers don't have to provide g++ alias or wrapper.
 DEFCC  := cc
@@ -312,7 +312,10 @@ ifdef LLAMA_DEBUG
 		MK_CPPFLAGS += -D_GLIBCXX_ASSERTIONS
 	endif
 else
-	MK_CPPFLAGS += -DNDEBUG
+	MK_CPPFLAGS   += -DNDEBUG
+	MK_CFLAGS     += -O3
+	MK_CXXFLAGS   += -O3
+	MK_NVCCFLAGS  += -O3
 endif

 ifdef LLAMA_SANITIZE_THREAD
@@ -1073,6 +1076,7 @@ clean:
 	rm -rvf src/*.o
 	rm -rvf tests/*.o
 	rm -rvf examples/*.o
+	rm -rvf common/*.o
 	rm -rvf *.a
 	rm -rvf *.dll
 	rm -rvf *.so
@@ -1087,6 +1091,7 @@ clean:
 	rm -vrf ggml/src/ggml-cuda/template-instances/*.o
 	rm -rvf $(BUILD_TARGETS)
 	rm -rvf $(TEST_TARGETS)
+	rm -rvf $(LEGACY_TARGETS)
 	find examples pocs -type f -name "*.o" -delete

 #
@@ -13,8 +13,9 @@ Inference of Meta's [LLaMA](https://arxiv.org/abs/2302.13971) model (and others)
 > [!IMPORTANT]
 [2024 Jun 12] Binaries have been renamed w/ a `llama-` prefix. `main` is now `llama-cli`, `server` is `llama-server`, etc (https://github.com/ggerganov/llama.cpp/pull/7809)

-### Recent API changes
+## Recent API changes

+- [2024 Jun 26] The source code and CMake build scripts have been restructured https://github.com/ggerganov/llama.cpp/pull/8006
 - [2024 Apr 21] `llama_token_to_piece` can now optionally render special tokens https://github.com/ggerganov/llama.cpp/pull/6807
 - [2024 Apr 4] State and session file functions reorganized under `llama_state_*` https://github.com/ggerganov/llama.cpp/pull/6341
 - [2024 Mar 26] Logits and embeddings API updated for compactness https://github.com/ggerganov/llama.cpp/pull/6122
@@ -23,9 +24,9 @@ Inference of Meta's [LLaMA](https://arxiv.org/abs/2302.13971) model (and others)
 - [2024 Mar 4] Embeddings API updated https://github.com/ggerganov/llama.cpp/pull/5796
 - [2024 Mar 3] `struct llama_context_params` https://github.com/ggerganov/llama.cpp/pull/5849

-### Hot topics
+## Hot topics

- **`convert.py` has been deprecated and moved to `examples/convert-legacy-llama.py`, please use `convert-hf-to-gguf.py`** https://github.com/ggerganov/llama.cpp/pull/7430
+- **`convert.py` has been deprecated and moved to `examples/convert_legacy_llama.py`, please use `convert_hf_to_gguf.py`** https://github.com/ggerganov/llama.cpp/pull/7430
 - Initial Flash-Attention support: https://github.com/ggerganov/llama.cpp/pull/5021
 - BPE pre-tokenization support has been added: https://github.com/ggerganov/llama.cpp/pull/6920
 - MoE memory layout has been updated - reconvert models for `mmap` support and regenerate `imatrix` https://github.com/ggerganov/llama.cpp/pull/6387
@@ -38,37 +39,6 @@ Inference of Meta's [LLaMA](https://arxiv.org/abs/2302.13971) model (and others)

 ----

-<details>
-  <summary>Table of Contents</summary>
-  <ol>
-    <li>
-      <a href="#description">Description</a>
-    </li>
-    <li>
-      <a href="#usage">Usage</a>
-      <ul>
-        <li><a href="#get-the-code">Get the Code</a></li>
-        <li><a href="#build">Build</a></li>
-        <li><a href="#blas-build">BLAS Build</a></li>
-        <li><a href="#prepare-and-quantize">Prepare and Quantize</a></li>
-        <li><a href="#run-the-quantized-model">Run the quantized model</a></li>
-        <li><a href="#memorydisk-requirements">Memory/Disk Requirements</a></li>
-        <li><a href="#quantization">Quantization</a></li>
-        <li><a href="#interactive-mode">Interactive mode</a></li>
-        <li><a href="#constrained-output-with-grammars">Constrained output with grammars</a></li>
-        <li><a href="#obtaining-and-using-the-facebook-llama-2-model">Obtaining and using the Facebook LLaMA 2 model</a></li>
-        <li><a href="#seminal-papers-and-background-on-the-models">Seminal papers and background on the models</a></li>
-        <li><a href="#perplexity-measuring-model-quality">Perplexity (measuring model quality)</a></li>
-        <li><a href="#android">Android</a></li>
-        <li><a href="#docker">Docker</a></li>
-      </ul>
-    </li>
-    <li><a href="#contributing">Contributing</a></li>
-    <li><a href="#coding-guidelines">Coding guidelines</a></li>
-    <li><a href="#docs">Docs</a></li>
-  </ol>
-</details>
-
 ## Description

 The main goal of `llama.cpp` is to enable LLM inference with minimal setup and state-of-the-art performance on a wide
@@ -86,14 +56,6 @@ Since its [inception](https://github.com/ggerganov/llama.cpp/issues/33#issuecomm
 improved significantly thanks to many contributions. It is the main playground for developing new features for the
 [ggml](https://github.com/ggerganov/ggml) library.

-**Supported platforms:**
-
- [X] Mac OS
- [X] Linux
- [X] Windows (via CMake)
- [X] Docker
- [X] FreeBSD
-
 **Supported models:**

 Typically finetunes of the base models below are supported as well.
@@ -107,6 +69,7 @@ Typically finetunes of the base models below are supported as well.
 - [X] [Falcon](https://huggingface.co/models?search=tiiuae/falcon)
 - [X] [Chinese LLaMA / Alpaca](https://github.com/ymcui/Chinese-LLaMA-Alpaca) and [Chinese LLaMA-2 / Alpaca-2](https://github.com/ymcui/Chinese-LLaMA-Alpaca-2)
 - [X] [Vigogne (French)](https://github.com/bofenghuang/vigogne)
+- [X] [BERT](https://github.com/ggerganov/llama.cpp/pull/5423)
 - [X] [Koala](https://bair.berkeley.edu/blog/2023/04/03/koala/)
 - [X] [Baichuan 1 & 2](https://huggingface.co/models?search=baichuan-inc/Baichuan) + [derivations](https://huggingface.co/hiyouga/baichuan-7b-sft)
 - [X] [Aquila 1 & 2](https://huggingface.co/models?search=BAAI/Aquila)
@@ -148,12 +111,6 @@ Typically finetunes of the base models below are supported as well.
 - [x] [Moondream](https://huggingface.co/vikhyatk/moondream2)
 - [x] [Bunny](https://github.com/BAAI-DCAI/Bunny)

-**HTTP server**
-
-[llama.cpp web server](./examples/server) is a lightweight [OpenAI API](https://github.com/openai/openai-openapi) compatible HTTP server that can be used to serve local models and easily connect them to existing clients.
-
-[simplechat](./examples/server/public_simplechat) is a simple chat client, which can be used to chat with the model exposed using above web server (use --path to point to simplechat), from a local web browser.
-
 **Bindings:**

 - Python: [abetlen/llama-cpp-python](https://github.com/abetlen/llama-cpp-python)
@@ -216,10 +173,16 @@ Unless otherwise noted these projects are open-source with permissive licensing:
 **Tools:**

 - [akx/ggify](https://github.com/akx/ggify) – download PyTorch models from HuggingFace Hub and convert them to GGML
+- [crashr/gppm](https://github.com/crashr/gppm) – launch llama.cpp instances utilizing NVIDIA Tesla P40 or P100 GPUs with reduced idle power consumption

---
+**Infrastructure:**

-Here is a typical run using LLaMA v2 13B on M2 Ultra:
+- [Paddler](https://github.com/distantmagic/paddler) - Stateful load balancer custom-tailored for llama.cpp
+
+## Demo
+
+<details>
+<summary>Typical run using LLaMA v2 13B on M2 Ultra</summary>

 ```
 $ make -j && ./llama-cli -m models/llama-13b-v2/ggml-model-q4_0.gguf -p "Building a website can be done in 10 simple steps:\nStep 1:" -n 400 -e
@@ -299,454 +262,85 @@ llama_print_timings:        eval time = 24513.59 ms /   399 runs   (   61.44 ms
 llama_print_timings:       total time = 25431.49 ms
 ```

+</details>
+
+<details>
+<summary>Demo of running both LLaMA-7B and whisper.cpp on a single M1 Pro MacBook</summary>
+
 And here is another demo of running both LLaMA-7B and [whisper.cpp](https://github.com/ggerganov/whisper.cpp) on a single M1 Pro MacBook:

 https://user-images.githubusercontent.com/1991296/224442907-7693d4be-acaa-4e01-8b4f-add84093ffff.mp4

+</details>
+
 ## Usage

 Here are the end-to-end binary build and model conversion steps for most supported models.

-### Get the Code
+### Basic usage
+
+Firstly, you need to get the binary. There are different methods that you can follow:
+- Method 1: Clone this repository and build locally, see [how to build](./docs/build.md)
+- Method 2: If you are using MacOS or Linux, you can install llama.cpp via [brew, flox or nix](./docs/install.md)
+- Method 3: Use a Docker image, see [documentation for Docker](./docs/docker.md)
+- Method 4: Download pre-built binary from [releases](https://github.com/ggerganov/llama.cpp/releases)
+
+You can run a basic completion using this command:

 ```bash
-git clone https://github.com/ggerganov/llama.cpp
-cd llama.cpp
+llama-cli -m your_model.gguf -p "I believe the meaning of life is" -n 128
+
+# Output:
+# I believe the meaning of life is to find your own truth and to live in accordance with it. For me, this means being true to myself and following my passions, even if they don't align with societal expectations. I think that's what I love about yoga – it's not just a physical practice, but a spiritual one too. It's about connecting with yourself, listening to your inner voice, and honoring your own unique journey.
 ```

-### Build
+See [this page](./examples/main/README.md) for a full list of parameters.

-In order to build llama.cpp you have four different options.
+### Conversation mode

- Using `make`:
-  - On Linux or MacOS:
-
-      ```bash
-      make
-      ```
-
-  - On Windows:
-
-    1. Download the latest fortran version of [w64devkit](https://github.com/skeeto/w64devkit/releases).
-    2. Extract `w64devkit` on your pc.
-    3. Run `w64devkit.exe`.
-    4. Use the `cd` command to reach the `llama.cpp` folder.
-    5. From here you can run:
-        ```bash
-        make
-        ```
-
-  - Notes:
-    - For faster compilation, add the `-j` argument to run multiple jobs in parallel. For example, `make -j 8` will run 8 jobs in parallel.
-    - For faster repeated compilation, install [ccache](https://ccache.dev/).
-    - For debug builds, run `make LLAMA_DEBUG=1`
-
- Using `CMake`:
-
-  ```bash
-  cmake -B build
-  cmake --build build --config Release
-  ```
-
-  **Notes**:
-
-    - For faster compilation, add the `-j` argument to run multiple jobs in parallel. For example, `cmake --build build --config Release -j 8` will run 8 jobs in parallel.
-    - For faster repeated compilation, install [ccache](https://ccache.dev/).
-    - For debug builds, there are two cases:
-
-      1. Single-config generators (e.g. default = `Unix Makefiles`; note that they just ignore the `--config` flag):
-
-      ```bash
-      cmake -B build -DCMAKE_BUILD_TYPE=Debug
-      cmake --build build
-      ```
-
-      2. Multi-config generators (`-G` param set to Visual Studio, XCode...):
-
-      ```bash
-      cmake -B build -G "Xcode"
-      cmake --build build --config Debug
-      ```
-
-   Using `gmake` (FreeBSD):
-
-    1. Install and activate [DRM in FreeBSD](https://wiki.freebsd.org/Graphics)
-    2. Add your user to **video** group
-    3. Install compilation dependencies.
-
-        ```bash
-        sudo pkg install gmake automake autoconf pkgconf llvm15 openblas
-
-        gmake CC=/usr/local/bin/clang15 CXX=/usr/local/bin/clang++15 -j4
-        ```
-
-### Homebrew
-
-On Mac and Linux, the homebrew package manager can be used via
-```
-brew install llama.cpp
-```
-The formula is automatically updated with new `llama.cpp` releases. More info: https://github.com/ggerganov/llama.cpp/discussions/7668
-
-### Nix
-
-On Mac and Linux, the Nix package manager can be used via
-```
-nix profile install nixpkgs#llama-cpp
-```
-For flake enabled installs.
-
-Or
-```
-nix-env --file '<nixpkgs>' --install --attr llama-cpp
-```
-For non-flake enabled installs.
-
-This expression is automatically updated within the [nixpkgs repo](https://github.com/NixOS/nixpkgs/blob/nixos-24.05/pkgs/by-name/ll/llama-cpp/package.nix#L164).
-
-#### Flox
-
-On Mac and Linux, Flox can be used to install llama.cpp within a Flox environment via
-```
-flox install llama-cpp
-```
-Flox follows the nixpkgs build of llama.cpp.
-
-### Metal Build
-
-On MacOS, Metal is enabled by default. Using Metal makes the computation run on the GPU.
-To disable the Metal build at compile time use the `GGML_NO_METAL=1` flag or the `GGML_METAL=OFF` cmake option.
-
-When built with Metal support, you can explicitly disable GPU inference with the `--n-gpu-layers|-ngl 0` command-line
-argument.
-
-### BLAS Build
-
-Building the program with BLAS support may lead to some performance improvements in prompt processing using batch sizes higher than 32 (the default is 512). Support with CPU-only BLAS implementations doesn't affect the normal generation performance. We may see generation performance improvements with GPU-involved BLAS implementations, e.g. cuBLAS, hipBLAS. There are currently several different BLAS implementations available for build and use:
-
- #### Accelerate Framework:
-
-  This is only available on Mac PCs and it's enabled by default. You can just build using the normal instructions.
-
- #### OpenBLAS:
-
-  This provides BLAS acceleration using only the CPU. Make sure to have OpenBLAS installed on your machine.
-
-  - Using `make`:
-    - On Linux:
-      ```bash
-      make GGML_OPENBLAS=1
-      ```
-
-    - On Windows:
-
-      1. Download the latest fortran version of [w64devkit](https://github.com/skeeto/w64devkit/releases).
-      2. Download the latest version of [OpenBLAS for Windows](https://github.com/xianyi/OpenBLAS/releases).
-      3. Extract `w64devkit` on your pc.
-      4. From the OpenBLAS zip that you just downloaded copy `libopenblas.a`, located inside the `lib` folder, inside `w64devkit\x86_64-w64-mingw32\lib`.
-      5. From the same OpenBLAS zip copy the content of the `include` folder inside `w64devkit\x86_64-w64-mingw32\include`.
-      6. Run `w64devkit.exe`.
-      7. Use the `cd` command to reach the `llama.cpp` folder.
-      8. From here you can run:
-
-          ```bash
-          make GGML_OPENBLAS=1
-          ```
-
-  - Using `CMake` on Linux:
-
-      ```bash
-      cmake -B build -DGGML_BLAS=ON -DGGML_BLAS_VENDOR=OpenBLAS
-      cmake --build build --config Release
-      ```
-
- #### BLIS
-
-  Check [BLIS.md](docs/BLIS.md) for more information.
-
- #### SYCL
-  SYCL is a higher-level programming model to improve programming productivity on various hardware accelerators.
-
-  llama.cpp based on SYCL is used to **support Intel GPU** (Data Center Max series, Flex series, Arc series, Built-in GPU and iGPU).
-
-  For detailed info, please refer to [llama.cpp for SYCL](README-sycl.md).
-
- #### Intel oneMKL
-  Building through oneAPI compilers will make avx_vnni instruction set available for intel processors that do not support avx512 and avx512_vnni. Please note that this build config **does not support Intel GPU**. For Intel GPU support, please refer to [llama.cpp for SYCL](./README-sycl.md).
-
-  - Using manual oneAPI installation:
-    By default, `GGML_BLAS_VENDOR` is set to `Generic`, so if you already sourced intel environment script and assign `-DGGML_BLAS=ON` in cmake, the mkl version of Blas will automatically been selected. Otherwise please install oneAPI and follow the below steps:
-      ```bash
-      source /opt/intel/oneapi/setvars.sh # You can skip this step if  in oneapi-basekit docker image, only required for manual installation
-      cmake -B build -DGGML_BLAS=ON -DGGML_BLAS_VENDOR=Intel10_64lp -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx -DGGML_NATIVE=ON
-      cmake --build build --config Release
-      ```
-
-  - Using oneAPI docker image:
-    If you do not want to source the environment vars and install oneAPI manually, you can also build the code using intel docker container: [oneAPI-basekit](https://hub.docker.com/r/intel/oneapi-basekit). Then, you can use the commands given above.
-
-  Check [Optimizing and Running LLaMA2 on Intel® CPU](https://www.intel.com/content/www/us/en/content-details/791610/optimizing-and-running-llama2-on-intel-cpu.html) for more information.
-
- #### CUDA
-
-  This provides GPU acceleration using the CUDA cores of your Nvidia GPU. Make sure to have the CUDA toolkit installed. You can download it from your Linux distro's package manager (e.g. `apt install nvidia-cuda-toolkit`) or from here: [CUDA Toolkit](https://developer.nvidia.com/cuda-downloads).
-
-  For Jetson user, if you have Jetson Orin, you can try this: [Offical Support](https://www.jetson-ai-lab.com/tutorial_text-generation.html). If you are using an old model(nano/TX2), need some additional operations before compiling.
-
-  - Using `make`:
-    ```bash
-    make GGML_CUDA=1
-    ```
-  - Using `CMake`:
-
-    ```bash
-    cmake -B build -DGGML_CUDA=ON
-    cmake --build build --config Release
-    ```
-
-  The environment variable [`CUDA_VISIBLE_DEVICES`](https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#env-vars) can be used to specify which GPU(s) will be used. The following compilation options are also available to tweak performance:
-
-  | Option                        | Legal values           | Default | Description                                                                                                                                                                                                                                                                             |
-  |-------------------------------|------------------------|---------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
-  | GGML_CUDA_FORCE_DMMV          | Boolean                | false   | Force the use of dequantization + matrix vector multiplication kernels instead of using kernels that do matrix vector multiplication on quantized data. By default the decision is made based on compute capability (MMVQ for 6.1/Pascal/GTX 1000 or higher). Does not affect k-quants. |
-  | GGML_CUDA_DMMV_X              | Positive integer >= 32 | 32      | Number of values in x direction processed by the CUDA dequantization + matrix vector multiplication kernel per iteration. Increasing this value can improve performance on fast GPUs. Power of 2 heavily recommended. Does not affect k-quants.                                         |
-  | GGML_CUDA_MMV_Y               | Positive integer       | 1       | Block size in y direction for the CUDA mul mat vec kernels. Increasing this value can improve performance on fast GPUs. Power of 2 recommended.                                                                                                                                         |
-  | GGML_CUDA_FORCE_MMQ           | Boolean                | false   | Force the use of custom matrix multiplication kernels for quantized models instead of FP16 cuBLAS even if there is no int8 tensor core implementation available (affects V100, RDNA3). MMQ kernels are enabled by default on GPUs with int8 tensor core support. With MMQ force enabled, speed for large batch sizes will be worse but VRAM consumption will be lower.                       |
-  | GGML_CUDA_FORCE_CUBLAS        | Boolean                | false   | Force the use of FP16 cuBLAS instead of custom matrix multiplication kernels for quantized models                                                                                                                                                                                       |
-  | GGML_CUDA_F16                 | Boolean                | false   | If enabled, use half-precision floating point arithmetic for the CUDA dequantization + mul mat vec kernels and for the q4_1 and q5_1 matrix matrix multiplication kernels. Can improve performance on relatively recent GPUs.                                                           |
-  | GGML_CUDA_KQUANTS_ITER        | 1 or 2                 | 2       | Number of values processed per iteration and per CUDA thread for Q2_K and Q6_K quantization formats. Setting this value to 1 can improve performance for slow GPUs.                                                                                                                     |
-  | GGML_CUDA_PEER_MAX_BATCH_SIZE | Positive integer       | 128     | Maximum batch size for which to enable peer access between multiple GPUs. Peer access requires either Linux or NVLink. When using NVLink enabling peer access for larger batch sizes is potentially beneficial.                                                                         |
-  | GGML_CUDA_FA_ALL_QUANTS       | Boolean                | false   | Compile support for all KV cache quantization type (combinations) for the FlashAttention CUDA kernels. More fine-grained control over KV cache size but compilation takes much longer.                                                                                                  |
-
- #### hipBLAS
-
-  This provides BLAS acceleration on HIP-supported AMD GPUs.
-  Make sure to have ROCm installed.
-  You can download it from your Linux distro's package manager or from here: [ROCm Quick Start (Linux)](https://rocm.docs.amd.com/projects/install-on-linux/en/latest/tutorial/quick-start.html#rocm-install-quick).
-
-  - Using `make`:
-    ```bash
-    make GGML_HIPBLAS=1
-    ```
-  - Using `CMake` for Linux (assuming a gfx1030-compatible AMD GPU):
-    ```bash
-    HIPCXX="$(hipconfig -l)/clang" HIP_PATH="$(hipconfig -R)" \
-        cmake -S . -B build -DGGML_HIPBLAS=ON -DAMDGPU_TARGETS=gfx1030 -DCMAKE_BUILD_TYPE=Release \
-        && cmake --build build --config Release -- -j 16
-    ```
-    On Linux it is also possible to use unified memory architecture (UMA) to share main memory between the CPU and integrated GPU by setting `-DGGML_HIP_UMA=ON`.
-    However, this hurts performance for non-integrated GPUs (but enables working with integrated GPUs).
-
-    Note that if you get the following error:
-    ```
-    clang: error: cannot find ROCm device library; provide its path via '--rocm-path' or '--rocm-device-lib-path', or pass '-nogpulib' to build without ROCm device library
-    ```
-    Try searching for a directory under `HIP_PATH` that contains the file
-    `oclc_abi_version_400.bc`. Then, add the following to the start of the
-    command: `HIP_DEVICE_LIB_PATH=<directory-you-just-found>`, so something
-    like:
-    ```bash
-    HIPCXX="$(hipconfig -l)/clang" HIP_PATH="$(hipconfig -p)" \
-    HIP_DEVICE_LIB_PATH=<directory-you-just-found> \
-        cmake -S . -B build -DGGML_HIPBLAS=ON -DAMDGPU_TARGETS=gfx1030 -DCMAKE_BUILD_TYPE=Release \
-        && cmake --build build -- -j 16
-    ```
-
-  - Using `make` (example for target gfx1030, build with 16 CPU threads):
-    ```bash
-    make -j16 GGML_HIPBLAS=1 GGML_HIP_UMA=1 AMDGPU_TARGETS=gfx1030
-    ```
-
-  - Using `CMake` for Windows (using x64 Native Tools Command Prompt for VS, and assuming a gfx1100-compatible AMD GPU):
-    ```bash
-    set PATH=%HIP_PATH%\bin;%PATH%
-    cmake -S . -B build -G Ninja -DAMDGPU_TARGETS=gfx1100 -DGGML_HIPBLAS=ON -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ -DCMAKE_BUILD_TYPE=Release
-    cmake --build build
-    ```
-    Make sure that `AMDGPU_TARGETS` is set to the GPU arch you want to compile for. The above example uses `gfx1100` that corresponds to Radeon RX 7900XTX/XT/GRE. You can find a list of targets [here](https://llvm.org/docs/AMDGPUUsage.html#processors)
-    Find your gpu version string by matching the most significant version information from `rocminfo | grep gfx | head -1 | awk '{print $2}'` with the list of processors, e.g. `gfx1035` maps to `gfx1030`.
-
-
-  The environment variable [`HIP_VISIBLE_DEVICES`](https://rocm.docs.amd.com/en/latest/understand/gpu_isolation.html#hip-visible-devices) can be used to specify which GPU(s) will be used.
-  If your GPU is not officially supported you can use the environment variable [`HSA_OVERRIDE_GFX_VERSION`] set to a similar GPU, for example 10.3.0 on RDNA2 (e.g. gfx1030, gfx1031, or gfx1035) or 11.0.0 on RDNA3.
-  The following compilation options are also available to tweak performance (yes, they refer to CUDA, not HIP, because it uses the same code as the cuBLAS version above):
-
-  | Option                 | Legal values           | Default | Description                                                                                                                                                                                                                                    |
-  |------------------------|------------------------|---------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
-  | GGML_CUDA_DMMV_X       | Positive integer >= 32 | 32      | Number of values in x direction processed by the HIP dequantization + matrix vector multiplication kernel per iteration. Increasing this value can improve performance on fast GPUs. Power of 2 heavily recommended. Does not affect k-quants. |
-  | GGML_CUDA_MMV_Y        | Positive integer       | 1       | Block size in y direction for the HIP mul mat vec kernels. Increasing this value can improve performance on fast GPUs. Power of 2 recommended. Does not affect k-quants.                                                                       |
-  | GGML_CUDA_KQUANTS_ITER | 1 or 2                 | 2       | Number of values processed per iteration and per HIP thread for Q2_K and Q6_K quantization formats. Setting this value to 1 can improve performance for slow GPUs.                                                                             |
-
- #### Vulkan
-
-  **With docker**:
-
-  You don't need to install Vulkan SDK. It will be installed inside the container.
-
-  ```sh
-  # Build the image
-  docker build -t llama-cpp-vulkan -f .devops/llama-cli-vulkan.Dockerfile .
-
-  # Then, use it:
-  docker run -it --rm -v "$(pwd):/app:Z" --device /dev/dri/renderD128:/dev/dri/renderD128 --device /dev/dri/card1:/dev/dri/card1 llama-cpp-vulkan -m "/app/models/YOUR_MODEL_FILE" -p "Building a website can be done in 10 simple steps:" -n 400 -e -ngl 33
-  ```
-
-  **Without docker**:
-
-  Firstly, you need to make sure you have installed [Vulkan SDK](https://vulkan.lunarg.com/doc/view/latest/linux/getting_started_ubuntu.html)
-
-  For example, on Ubuntu 22.04 (jammy), use the command below:
-
-  ```bash
-  wget -qO - https://packages.lunarg.com/lunarg-signing-key-pub.asc | apt-key add -
-  wget -qO /etc/apt/sources.list.d/lunarg-vulkan-jammy.list https://packages.lunarg.com/vulkan/lunarg-vulkan-jammy.list
-  apt update -y
-  apt-get install -y vulkan-sdk
-  # To verify the installation, use the command below:
-  vulkaninfo
-  ```
-
-  Alternatively your package manager might be able to provide the appropriate libraries.
-  For example for Ubuntu 22.04 you can install `libvulkan-dev` instead.
-  For Fedora 40, you can install `vulkan-devel`, `glslc` and `glslang` packages.
-
-  Then, build llama.cpp using the cmake command below:
-
-  ```bash
-  cmake -B build -DGGML_VULKAN=1
-  cmake --build build --config Release
-  # Test the output binary (with "-ngl 33" to offload all layers to GPU)
-  ./bin/llama-cli -m "PATH_TO_MODEL" -p "Hi you how are you" -n 50 -e -ngl 33 -t 4
-
-  # You should see in the output, ggml_vulkan detected your GPU. For example:
-  # ggml_vulkan: Using Intel(R) Graphics (ADL GT2) | uma: 1 | fp16: 1 | warp size: 32
-  ```
-
-### Prepare and Quantize
-
-> [!NOTE]
-> You can use the [GGUF-my-repo](https://huggingface.co/spaces/ggml-org/gguf-my-repo) space on Hugging Face to quantise your model weights without any setup too. It is synced from `llama.cpp` main every 6 hours.
-
-To obtain the official LLaMA 2 weights please see the <a href="#obtaining-and-using-the-facebook-llama-2-model">Obtaining and using the Facebook LLaMA 2 model</a> section. There is also a large selection of pre-quantized `gguf` models available on Hugging Face.
-
-Note: `convert.py` has been moved to `examples/convert-legacy-llama.py` and shouldn't be used for anything other than `Llama/Llama2/Mistral` models and their derivatives.
-It does not support LLaMA 3, you can use `convert-hf-to-gguf.py` with LLaMA 3 downloaded from Hugging Face.
+If you want a more ChatGPT-like experience, you can run in conversation mode by passing `-cnv` as a parameter:

 ```bash
-# obtain the official LLaMA model weights and place them in ./models
-ls ./models
-llama-2-7b tokenizer_checklist.chk tokenizer.model
-# [Optional] for models using BPE tokenizers
-ls ./models
-<folder containing weights and tokenizer json> vocab.json
-# [Optional] for PyTorch .bin models like Mistral-7B
-ls ./models
-<folder containing weights and tokenizer json>
+llama-cli -m your_model.gguf -p "You are a helpful assistant" -cnv

-# install Python dependencies
-python3 -m pip install -r requirements.txt
-
-# convert the model to ggml FP16 format
-python3 convert-hf-to-gguf.py models/mymodel/
-
-# quantize the model to 4-bits (using Q4_K_M method)
-./llama-quantize ./models/mymodel/ggml-model-f16.gguf ./models/mymodel/ggml-model-Q4_K_M.gguf Q4_K_M
-
-# update the gguf filetype to current version if older version is now unsupported
-./llama-quantize ./models/mymodel/ggml-model-Q4_K_M.gguf ./models/mymodel/ggml-model-Q4_K_M-v2.gguf COPY
+# Output:
+# > hi, who are you?
+# Hi there! I'm your helpful assistant! I'm an AI-powered chatbot designed to assist and provide information to users like you. I'm here to help answer your questions, provide guidance, and offer support on a wide range of topics. I'm a friendly and knowledgeable AI, and I'm always happy to help with anything you need. What's on your mind, and how can I assist you today?
+#
+# > what is 1+1?
+# Easy peasy! The answer to 1+1 is... 2!
 ```

-### Run the quantized model
+By default, the chat template will be taken from the input model. If you want to use another chat template, pass `--chat-template NAME` as a parameter. See the list of [supported templates](https://github.com/ggerganov/llama.cpp/wiki/Templates-supported-by-llama_chat_apply_template)

 ```bash
-# start inference on a gguf model
-./llama-cli -m ./models/mymodel/ggml-model-Q4_K_M.gguf -n 128
+./llama-cli -m your_model.gguf -p "You are a helpful assistant" -cnv --chat-template chatml
 ```

-When running the larger models, make sure you have enough disk space to store all the intermediate files.
+You can also use your own template via in-prefix, in-suffix and reverse-prompt parameters:

-### Running on Windows with prebuilt binaries
-
-You will find prebuilt Windows binaries on the release page.
-
-Simply download and extract the latest zip package of choice: (e.g. `llama-b1380-bin-win-avx2-x64.zip`)
-
-From the unzipped folder, open a terminal/cmd window here and place a pre-converted `.gguf` model file. Test out the main example like so:
-
-```
-.\main -m llama-2-7b.Q4_0.gguf -n 128
+```bash
+./llama-cli -m your_model.gguf -p "You are a helpful assistant" -cnv --in-prefix 'User: ' --reverse-prompt 'User:'
 ```

-### Memory/Disk Requirements
+### Web server

-As the models are currently fully loaded into memory, you will need adequate disk space to save them and sufficient RAM to load them. At the moment, memory and disk requirements are the same.
+[llama.cpp web server](./examples/server/README.md) is a lightweight [OpenAI API](https://github.com/openai/openai-openapi) compatible HTTP server that can be used to serve local models and easily connect them to existing clients.

-| Model | Original size | Quantized size (Q4_0) |
-|------:|--------------:|----------------------:|
-|    7B |         13 GB |                3.9 GB |
-|   13B |         24 GB |                7.8 GB |
-|   30B |         60 GB |               19.5 GB |
-|   65B |        120 GB |               38.5 GB |
+Example usage:

-### Quantization
+```bash
+./llama-server -m your_model.gguf --port 8080

-Several quantization methods are supported. They differ in the resulting model disk size and inference speed.
-
-*(outdated)*
-
-| Model | Measure      |    F16 |   Q4_0 |   Q4_1 |   Q5_0 |   Q5_1 |   Q8_0 |
-|------:|--------------|-------:|-------:|-------:|-------:|-------:|-------:|
-|    7B | perplexity   | 5.9066 | 6.1565 | 6.0912 | 5.9862 | 5.9481 | 5.9070 |
-|    7B | file size    |  13.0G |   3.5G |   3.9G |   4.3G |   4.7G |   6.7G |
-|    7B | ms/tok @ 4th |    127 |     55 |     54 |     76 |     83 |     72 |
-|    7B | ms/tok @ 8th |    122 |     43 |     45 |     52 |     56 |     67 |
-|    7B | bits/weight  |   16.0 |    4.5 |    5.0 |    5.5 |    6.0 |    8.5 |
-|   13B | perplexity   | 5.2543 | 5.3860 | 5.3608 | 5.2856 | 5.2706 | 5.2548 |
-|   13B | file size    |  25.0G |   6.8G |   7.6G |   8.3G |   9.1G |    13G |
-|   13B | ms/tok @ 4th |      - |    103 |    105 |    148 |    160 |    131 |
-|   13B | ms/tok @ 8th |      - |     73 |     82 |     98 |    105 |    128 |
-|   13B | bits/weight  |   16.0 |    4.5 |    5.0 |    5.5 |    6.0 |    8.5 |
-
- [k-quants](https://github.com/ggerganov/llama.cpp/pull/1684)
- recent k-quants improvements and new i-quants
-  - [#2707](https://github.com/ggerganov/llama.cpp/pull/2707)
-  - [#2807](https://github.com/ggerganov/llama.cpp/pull/2807)
-  - [#4773 - 2-bit i-quants (inference)](https://github.com/ggerganov/llama.cpp/pull/4773)
-  - [#4856 - 2-bit i-quants (inference)](https://github.com/ggerganov/llama.cpp/pull/4856)
-  - [#4861 - importance matrix](https://github.com/ggerganov/llama.cpp/pull/4861)
-  - [#4872 - MoE models](https://github.com/ggerganov/llama.cpp/pull/4872)
-  - [#4897 - 2-bit quantization](https://github.com/ggerganov/llama.cpp/pull/4897)
-  - [#4930 - imatrix for all k-quants](https://github.com/ggerganov/llama.cpp/pull/4930)
-  - [#4951 - imatrix on the GPU](https://github.com/ggerganov/llama.cpp/pull/4957)
-  - [#4969 - imatrix for legacy quants](https://github.com/ggerganov/llama.cpp/pull/4969)
-  - [#4996 - k-qunats tuning](https://github.com/ggerganov/llama.cpp/pull/4996)
-  - [#5060 - Q3_K_XS](https://github.com/ggerganov/llama.cpp/pull/5060)
-  - [#5196 - 3-bit i-quants](https://github.com/ggerganov/llama.cpp/pull/5196)
-  - [quantization tuning](https://github.com/ggerganov/llama.cpp/pull/5320), [another one](https://github.com/ggerganov/llama.cpp/pull/5334), and [another one](https://github.com/ggerganov/llama.cpp/pull/5361)
-
-### Perplexity (measuring model quality)
-
-You can use the `perplexity` example to measure perplexity over a given prompt (lower perplexity is better).
-For more information, see [https://huggingface.co/docs/transformers/perplexity](https://huggingface.co/docs/transformers/perplexity).
-
-The perplexity measurements in table above are done against the `wikitext2` test dataset (https://paperswithcode.com/dataset/wikitext-2), with context length of 512.
-The time per token is measured on a MacBook M1 Pro 32GB RAM using 4 and 8 threads.
-
-#### How to run
-
-1. Download/extract: https://huggingface.co/datasets/ggml-org/ci/resolve/main/wikitext-2-raw-v1.zip
-2. Run `./llama-perplexity -m models/7B/ggml-model-q4_0.gguf -f wiki.test.raw`
-3. Output:
+# Basic web UI can be accessed via browser: http://localhost:8080
+# Chat completion endpoint: http://localhost:8080/v1/chat/completions
 ```
-perplexity : calculating perplexity over 655 chunks
-24.43 seconds per pass - ETA 4.45 hours
-[1]4.5970,[2]5.1807,[3]6.0382,...
-```
-And after 4.45 hours, you will have the final perplexity.

 ### Interactive mode

-If you want a more ChatGPT-like experience, you can run in interactive mode by passing `-i` as a parameter.
+> [!NOTE]
+> If you prefer basic usage, please consider using conversation mode instead of interactive mode
+
 In this mode, you can always interrupt generation by pressing Ctrl+C and entering one or more lines of text, which will be converted into tokens and appended to the current context. You can also specify a *reverse prompt* with the parameter `-r "reverse prompt string"`. This will result in user input being prompted whenever the exact tokens of the reverse prompt string are encountered in the generation. A typical use is to use a prompt that makes LLaMA emulate a chat between multiple users, say Alice and Bob, and pass `-r "Alice:"`.

 Here is an example of a few-shot interaction, invoked with the command
@@ -797,18 +391,70 @@ The `grammars/` folder contains a handful of sample grammars. To write your own,

 For authoring more complex JSON grammars, you can also check out https://grammar.intrinsiclabs.ai/, a browser app that lets you write TypeScript interfaces which it compiles to GBNF grammars that you can save for local use. Note that the app is built and maintained by members of the community, please file any issues or FRs on [its repo](http://github.com/intrinsiclabsai/gbnfgen) and not this one.

-### Obtaining and using the Facebook LLaMA 2 model
+## Build

- Refer to [Facebook's LLaMA download page](https://ai.meta.com/resources/models-and-libraries/llama-downloads/) if you want to access the model data.
- Alternatively, if you want to save time and space, you can download already converted and quantized models from [TheBloke](https://huggingface.co/TheBloke), including:
-  - [LLaMA 2 7B base](https://huggingface.co/TheBloke/Llama-2-7B-GGUF)
-  - [LLaMA 2 13B base](https://huggingface.co/TheBloke/Llama-2-13B-GGUF)
-  - [LLaMA 2 70B base](https://huggingface.co/TheBloke/Llama-2-70B-GGUF)
-  - [LLaMA 2 7B chat](https://huggingface.co/TheBloke/Llama-2-7B-chat-GGUF)
-  - [LLaMA 2 13B chat](https://huggingface.co/TheBloke/Llama-2-13B-chat-GGUF)
-  - [LLaMA 2 70B chat](https://huggingface.co/TheBloke/Llama-2-70B-chat-GGUF)
+Please refer to [Build llama.cpp locally](./docs/build.md)

-### Seminal papers and background on the models
+## Supported backends
+
+| Backend | Target devices |
+| --- | --- |
+| [Metal](./docs/build.md#metal-build) | Apple Silicon |
+| [BLAS](./docs/build.md#blas-build) | All |
+| [BLIS](./docs/backend/BLIS.md) | All |
+| [SYCL](./docs/backend/SYCL.md) | Intel and Nvidia GPU |
+| [CUDA](./docs/build.md#cuda) | Nvidia GPU |
+| [hipBLAS](./docs/build.md#hipblas) | AMD GPU |
+| [Vulkan](./docs/build.md#vulkan) | GPU |
+
+## Tools
+
+### Prepare and Quantize
+
+> [!NOTE]
+> You can use the [GGUF-my-repo](https://huggingface.co/spaces/ggml-org/gguf-my-repo) space on Hugging Face to quantise your model weights without any setup too. It is synced from `llama.cpp` main every 6 hours.
+
+To obtain the official LLaMA 2 weights please see the <a href="#obtaining-and-using-the-facebook-llama-2-model">Obtaining and using the Facebook LLaMA 2 model</a> section. There is also a large selection of pre-quantized `gguf` models available on Hugging Face.
+
+Note: `convert.py` has been moved to `examples/convert_legacy_llama.py` and shouldn't be used for anything other than `Llama/Llama2/Mistral` models and their derivatives.
+It does not support LLaMA 3, you can use `convert_hf_to_gguf.py` with LLaMA 3 downloaded from Hugging Face.
+
+To learn more about quantizing model, [read this documentation](./examples/quantize/README.md)
+
+### Perplexity (measuring model quality)
+
+You can use the `perplexity` example to measure perplexity over a given prompt (lower perplexity is better).
+For more information, see [https://huggingface.co/docs/transformers/perplexity](https://huggingface.co/docs/transformers/perplexity).
+
+To learn more how to measure perplexity using llama.cpp, [read this documentation](./examples/perplexity/README.md)
+
+## Contributing
+
+- Contributors can open PRs
+- Collaborators can push to branches in the `llama.cpp` repo and merge PRs into the `master` branch
+- Collaborators will be invited based on contributions
+- Any help with managing issues and PRs is very appreciated!
+- See [good first issues](https://github.com/ggerganov/llama.cpp/issues?q=is%3Aissue+is%3Aopen+label%3A%22good+first+issue%22) for tasks suitable for first contributions
+- Read the [CONTRIBUTING.md](CONTRIBUTING.md) for more information
+- Make sure to read this: [Inference at the edge](https://github.com/ggerganov/llama.cpp/discussions/205)
+- A bit of backstory for those who are interested: [Changelog podcast](https://changelog.com/podcast/532)
+
+## Other documentations
+
+- [main (cli)](./examples/main/README.md)
+- [server](./examples/server/README.md)
+- [jeopardy](./examples/jeopardy/README.md)
+- [GBNF grammars](./grammars/README.md)
+
+**Development documentations**
+
+- [How to build](./docs/build.md)
+- [Running on Docker](./docs/docker.md)
+- [Build on Android](./docs/android.md)
+- [Performance troubleshooting](./docs/token_generation_performance_tips.md)
+- [GGML tips & tricks](https://github.com/ggerganov/llama.cpp/wiki/GGML-Tips-&-Tricks)
+
+**Seminal papers and background on the models**

 If your issue is with model generation quality, then please at least scan the following links and papers to understand the limitations of LLaMA models. This is especially important when choosing an appropriate model size and appreciating both the significant and subtle differences between LLaMA models and ChatGPT:
 - LLaMA:
@@ -819,178 +465,3 @@ If your issue is with model generation quality, then please at least scan the fo
 - GPT-3.5 / InstructGPT / ChatGPT:
    - [Aligning language models to follow instructions](https://openai.com/research/instruction-following)
    - [Training language models to follow instructions with human feedback](https://arxiv.org/abs/2203.02155)
-
-### Android
-
-#### Build on Android using Termux
-[Termux](https://github.com/termux/termux-app#installation) is a method to execute `llama.cpp` on an Android device (no root required).
-```
-apt update && apt upgrade -y
-apt install git make cmake
-```
-
-It's recommended to move your model inside the `~/` directory for best performance:
-```
-cd storage/downloads
-mv model.gguf ~/
-```
-
-[Get the code](https://github.com/ggerganov/llama.cpp#get-the-code) & [follow the Linux build instructions](https://github.com/ggerganov/llama.cpp#build) to build `llama.cpp`.
-
-#### Building the Project using Android NDK
-Obtain the [Android NDK](https://developer.android.com/ndk) and then build with CMake.
-
-Execute the following commands on your computer to avoid downloading the NDK to your mobile. Alternatively, you can also do this in Termux:
-```
-$ mkdir build-android
-$ cd build-android
-$ export NDK=<your_ndk_directory>
-$ cmake -DCMAKE_TOOLCHAIN_FILE=$NDK/build/cmake/android.toolchain.cmake -DANDROID_ABI=arm64-v8a -DANDROID_PLATFORM=android-23 -DCMAKE_C_FLAGS=-march=armv8.4a+dotprod ..
-$ make
-```
-
-Install [termux](https://github.com/termux/termux-app#installation) on your device and run `termux-setup-storage` to get access to your SD card (if Android 11+ then run the command twice).
-
-Finally, copy these built `llama` binaries and the model file to your device storage. Because the file permissions in the Android sdcard cannot be changed, you can copy the executable files to the `/data/data/com.termux/files/home/bin` path, and then execute the following commands in Termux to add executable permission:
-
-(Assumed that you have pushed the built executable files to the /sdcard/llama.cpp/bin path using `adb push`)
-```
-$cp -r /sdcard/llama.cpp/bin /data/data/com.termux/files/home/
-$cd /data/data/com.termux/files/home/bin
-$chmod +x ./*
-```
-
-Download model [llama-2-7b-chat.Q4_K_M.gguf](https://huggingface.co/TheBloke/Llama-2-7B-Chat-GGUF/blob/main/llama-2-7b-chat.Q4_K_M.gguf), and push it to `/sdcard/llama.cpp/`, then move it to `/data/data/com.termux/files/home/model/`
-
-```
-$mv /sdcard/llama.cpp/llama-2-7b-chat.Q4_K_M.gguf /data/data/com.termux/files/home/model/
-```
-
-Now, you can start chatting:
-```
-$cd /data/data/com.termux/files/home/bin
-$./llama-cli -m ../model/llama-2-7b-chat.Q4_K_M.gguf -n 128 -cml
-```
-
-Here's a demo of an interactive session running on Pixel 5 phone:
-
-https://user-images.githubusercontent.com/271616/225014776-1d567049-ad71-4ef2-b050-55b0b3b9274c.mp4
-
-### Docker
-
-#### Prerequisites
-* Docker must be installed and running on your system.
-* Create a folder to store big models & intermediate files (ex. /llama/models)
-
-#### Images
-We have three Docker images available for this project:
-
-1. `ghcr.io/ggerganov/llama.cpp:full`: This image includes both the main executable file and the tools to convert LLaMA models into ggml and convert into 4-bit quantization. (platforms: `linux/amd64`, `linux/arm64`)
-2. `ghcr.io/ggerganov/llama.cpp:light`: This image only includes the main executable file. (platforms: `linux/amd64`, `linux/arm64`)
-3. `ghcr.io/ggerganov/llama.cpp:server`: This image only includes the server executable file. (platforms: `linux/amd64`, `linux/arm64`)
-
-Additionally, there the following images, similar to the above:
-
- `ghcr.io/ggerganov/llama.cpp:full-cuda`: Same as `full` but compiled with CUDA support. (platforms: `linux/amd64`)
- `ghcr.io/ggerganov/llama.cpp:light-cuda`: Same as `light` but compiled with CUDA support. (platforms: `linux/amd64`)
- `ghcr.io/ggerganov/llama.cpp:server-cuda`: Same as `server` but compiled with CUDA support. (platforms: `linux/amd64`)
- `ghcr.io/ggerganov/llama.cpp:full-rocm`: Same as `full` but compiled with ROCm support. (platforms: `linux/amd64`, `linux/arm64`)
- `ghcr.io/ggerganov/llama.cpp:light-rocm`: Same as `light` but compiled with ROCm support. (platforms: `linux/amd64`, `linux/arm64`)
- `ghcr.io/ggerganov/llama.cpp:server-rocm`: Same as `server` but compiled with ROCm support. (platforms: `linux/amd64`, `linux/arm64`)
-
-The GPU enabled images are not currently tested by CI beyond being built. They are not built with any variation from the ones in the Dockerfiles defined in [.devops/](.devops/) and the GitHub Action defined in [.github/workflows/docker.yml](.github/workflows/docker.yml). If you need different settings (for example, a different CUDA or ROCm library, you'll need to build the images locally for now).
-
-#### Usage
-
-The easiest way to download the models, convert them to ggml and optimize them is with the --all-in-one command which includes the full docker image.
-
-Replace `/path/to/models` below with the actual path where you downloaded the models.
-
-```bash
-docker run -v /path/to/models:/models ghcr.io/ggerganov/llama.cpp:full --all-in-one "/models/" 7B
-```
-
-On completion, you are ready to play!
-
-```bash
-docker run -v /path/to/models:/models ghcr.io/ggerganov/llama.cpp:full --run -m /models/7B/ggml-model-q4_0.gguf -p "Building a website can be done in 10 simple steps:" -n 512
-```
-
-or with a light image:
-
-```bash
-docker run -v /path/to/models:/models ghcr.io/ggerganov/llama.cpp:light -m /models/7B/ggml-model-q4_0.gguf -p "Building a website can be done in 10 simple steps:" -n 512
-```
-
-or with a server image:
-
-```bash
-docker run -v /path/to/models:/models -p 8000:8000 ghcr.io/ggerganov/llama.cpp:server -m /models/7B/ggml-model-q4_0.gguf --port 8000 --host 0.0.0.0 -n 512
-```
-
-### Docker With CUDA
-
-Assuming one has the [nvidia-container-toolkit](https://github.com/NVIDIA/nvidia-container-toolkit) properly installed on Linux, or is using a GPU enabled cloud, `cuBLAS` should be accessible inside the container.
-
-#### Building Locally
-
-```bash
-docker build -t local/llama.cpp:full-cuda -f .devops/full-cuda.Dockerfile .
-docker build -t local/llama.cpp:light-cuda -f .devops/llama-cli-cuda.Dockerfile .
-docker build -t local/llama.cpp:server-cuda -f .devops/llama-server-cuda.Dockerfile .
-```
-
-You may want to pass in some different `ARGS`, depending on the CUDA environment supported by your container host, as well as the GPU architecture.
-
-The defaults are:
-
- `CUDA_VERSION` set to `11.7.1`
- `CUDA_DOCKER_ARCH` set to `all`
-
-The resulting images, are essentially the same as the non-CUDA images:
-
-1. `local/llama.cpp:full-cuda`: This image includes both the main executable file and the tools to convert LLaMA models into ggml and convert into 4-bit quantization.
-2. `local/llama.cpp:light-cuda`: This image only includes the main executable file.
-3. `local/llama.cpp:server-cuda`: This image only includes the server executable file.
-
-#### Usage
-
-After building locally, Usage is similar to the non-CUDA examples, but you'll need to add the `--gpus` flag. You will also want to use the `--n-gpu-layers` flag.
-
-```bash
-docker run --gpus all -v /path/to/models:/models local/llama.cpp:full-cuda --run -m /models/7B/ggml-model-q4_0.gguf -p "Building a website can be done in 10 simple steps:" -n 512 --n-gpu-layers 1
-docker run --gpus all -v /path/to/models:/models local/llama.cpp:light-cuda -m /models/7B/ggml-model-q4_0.gguf -p "Building a website can be done in 10 simple steps:" -n 512 --n-gpu-layers 1
-docker run --gpus all -v /path/to/models:/models local/llama.cpp:server-cuda -m /models/7B/ggml-model-q4_0.gguf --port 8000 --host 0.0.0.0 -n 512 --n-gpu-layers 1
-```
-
-### Contributing
-
- Contributors can open PRs
- Collaborators can push to branches in the `llama.cpp` repo and merge PRs into the `master` branch
- Collaborators will be invited based on contributions
- Any help with managing issues and PRs is very appreciated!
- Make sure to read this: [Inference at the edge](https://github.com/ggerganov/llama.cpp/discussions/205)
- A bit of backstory for those who are interested: [Changelog podcast](https://changelog.com/podcast/532)
-
-### Coding guidelines
-
- Avoid adding third-party dependencies, extra files, extra headers, etc.
- Always consider cross-compatibility with other operating systems and architectures
- Avoid fancy looking modern STL constructs, use basic `for` loops, avoid templates, keep it simple
- There are no strict rules for the code style, but try to follow the patterns in the code (indentation, spaces, etc.). Vertical alignment makes things more readable and easier to batch edit
- Clean-up any trailing whitespaces, use 4 spaces for indentation, brackets on the same line, `void * ptr`, `int & a`
- See [good first issues](https://github.com/ggerganov/llama.cpp/issues?q=is%3Aissue+is%3Aopen+label%3A%22good+first+issue%22) for tasks suitable for first contributions
- Tensors store data in row-major order. We refer to dimension 0 as columns, 1 as rows, 2 as matrices
- Matrix multiplication is unconventional: [`C = ggml_mul_mat(ctx, A, B)`](https://github.com/ggerganov/llama.cpp/blob/880e352277fc017df4d5794f0c21c44e1eae2b84/ggml.h#L1058-L1064) means $C^T = A B^T \Leftrightarrow C = B A^T.$
-
-![matmul](media/matmul.png)
-
-### Docs
-
- [main (cli)](./examples/main/README.md)
- [server](./examples/server/README.md)
- [jeopardy](./examples/jeopardy/README.md)
- [BLIS](./docs/BLIS.md)
- [Performance troubleshooting](./docs/token_generation_performance_tips.md)
- [GGML tips & tricks](https://github.com/ggerganov/llama.cpp/wiki/GGML-Tips-&-Tricks)
- [GBNF grammars](./grammars/README.md)
@@ -287,7 +287,7 @@ function gg_run_open_llama_7b_v2 {
    (time cmake -DCMAKE_BUILD_TYPE=Release ${CMAKE_EXTRA} -DGGML_CUDA=1 .. ) 2>&1 | tee -a $OUT/${ci}-cmake.log
    (time make -j                                                           ) 2>&1 | tee -a $OUT/${ci}-make.log

-    python3 ../examples/convert-legacy-llama.py ${path_models} --outfile ${path_models}/ggml-model-f16.gguf
+    python3 ../examples/convert_legacy_llama.py ${path_models} --outfile ${path_models}/ggml-model-f16.gguf

    model_f16="${path_models}/ggml-model-f16.gguf"
    model_q8_0="${path_models}/ggml-model-q8_0.gguf"
@@ -421,7 +421,7 @@ function gg_run_pythia_1_4b {
    (time cmake -DCMAKE_BUILD_TYPE=Release ${CMAKE_EXTRA} .. ) 2>&1 | tee -a $OUT/${ci}-cmake.log
    (time make -j                                            ) 2>&1 | tee -a $OUT/${ci}-make.log

-    python3 ../convert-hf-to-gguf.py ${path_models} --outfile ${path_models}/ggml-model-f16.gguf
+    python3 ../convert_hf_to_gguf.py ${path_models} --outfile ${path_models}/ggml-model-f16.gguf

    model_f16="${path_models}/ggml-model-f16.gguf"
    model_q8_0="${path_models}/ggml-model-q8_0.gguf"
@@ -553,7 +553,7 @@ function gg_run_pythia_2_8b {
    (time cmake -DCMAKE_BUILD_TYPE=Release ${CMAKE_EXTRA} -DGGML_CUDA=1 .. ) 2>&1 | tee -a $OUT/${ci}-cmake.log
    (time make -j                                                           ) 2>&1 | tee -a $OUT/${ci}-make.log

-    python3 ../convert-hf-to-gguf.py ${path_models} --outfile ${path_models}/ggml-model-f16.gguf
+    python3 ../convert_hf_to_gguf.py ${path_models} --outfile ${path_models}/ggml-model-f16.gguf

    model_f16="${path_models}/ggml-model-f16.gguf"
    model_q8_0="${path_models}/ggml-model-q8_0.gguf"
@@ -688,7 +688,7 @@ function gg_run_embd_bge_small {
    (time cmake -DCMAKE_BUILD_TYPE=Release ${CMAKE_EXTRA} .. ) 2>&1 | tee -a $OUT/${ci}-cmake.log
    (time make -j                                            ) 2>&1 | tee -a $OUT/${ci}-make.log

-    python3 ../convert-hf-to-gguf.py ${path_models} --outfile ${path_models}/ggml-model-f16.gguf
+    python3 ../convert_hf_to_gguf.py ${path_models} --outfile ${path_models}/ggml-model-f16.gguf

    model_f16="${path_models}/ggml-model-f16.gguf"
    model_q8_0="${path_models}/ggml-model-q8_0.gguf"
@@ -190,6 +190,12 @@ int32_t cpu_get_num_math() {
 // CLI argument parsing
 //

+void gpt_params_handle_hf_token(gpt_params & params) {
+    if (params.hf_token.empty() && std::getenv("HF_TOKEN")) {
+        params.hf_token = std::getenv("HF_TOKEN");
+    }
+}
+
 void gpt_params_handle_model_default(gpt_params & params) {
    if (!params.hf_repo.empty()) {
        // short-hand to avoid specifying --hf-file -> default it to --model
@@ -237,6 +243,8 @@ bool gpt_params_parse_ex(int argc, char ** argv, gpt_params & params) {

    gpt_params_handle_model_default(params);

+    gpt_params_handle_hf_token(params);
+
    if (params.escape) {
        string_process_escapes(params.prompt);
        string_process_escapes(params.input_prefix);
@@ -472,6 +480,14 @@ bool gpt_params_find_arg(int argc, char ** argv, const std::string & arg, gpt_pa
        else { invalid_param = true; }
        return true;
    }
+    if (arg == "--attention") {
+        CHECK_ARG
+        std::string value(argv[i]);
+        /**/ if (value == "causal") { params.attention_type = LLAMA_ATTENTION_TYPE_CAUSAL; }
+        else if (value == "non-causal") { params.attention_type = LLAMA_ATTENTION_TYPE_NON_CAUSAL; }
+        else { invalid_param = true; }
+        return true;
+    }
    if (arg == "--defrag-thold" || arg == "-dt") {
        CHECK_ARG
        params.defrag_thold = std::stof(argv[i]);
@@ -644,6 +660,14 @@ bool gpt_params_find_arg(int argc, char ** argv, const std::string & arg, gpt_pa
        params.model_url = argv[i];
        return true;
    }
+    if (arg == "-hft" || arg == "--hf-token") {
+        if (++i >= argc) {
+          invalid_param = true;
+          return true;
+        }
+        params.hf_token = argv[i];
+        return true;
+    }
    if (arg == "-hfr" || arg == "--hf-repo") {
        CHECK_ARG
        params.hf_repo = argv[i];
@@ -757,7 +781,7 @@ bool gpt_params_find_arg(int argc, char ** argv, const std::string & arg, gpt_pa
        params.cache_type_v = argv[++i];
        return true;
    }
-    if (arg == "--multiline-input") {
+    if (arg == "-mli" || arg == "--multiline-input") {
        params.multiline_input = true;
        return true;
    }
@@ -1014,16 +1038,23 @@ bool gpt_params_find_arg(int argc, char ** argv, const std::string & arg, gpt_pa
    }
    if (arg == "--in-prefix-bos") {
        params.input_prefix_bos = true;
+        params.enable_chat_template = false;
        return true;
    }
    if (arg == "--in-prefix") {
        CHECK_ARG
        params.input_prefix = argv[i];
+        params.enable_chat_template = false;
        return true;
    }
    if (arg == "--in-suffix") {
        CHECK_ARG
        params.input_suffix = argv[i];
+        params.enable_chat_template = false;
+        return true;
+    }
+    if (arg == "--spm-infill") {
+        params.spm_infill = true;
        return true;
    }
    if (arg == "--grammar") {
@@ -1387,7 +1418,9 @@ void gpt_params_print_usage(int /*argc*/, char ** argv, const gpt_params & param
    options.push_back({ "*",           "       --keep N",               "number of tokens to keep from the initial prompt (default: %d, -1 = all)", params.n_keep });
    options.push_back({ "*",           "       --chunks N",             "max number of chunks to process (default: %d, -1 = all)", params.n_chunks });
    options.push_back({ "*",           "-fa,   --flash-attn",           "enable Flash Attention (default: %s)", params.flash_attn ? "enabled" : "disabled" });
-    options.push_back({ "*",           "-p,    --prompt PROMPT",        "prompt to start generation with (default: '%s')", params.prompt.c_str() });
+    options.push_back({ "*",           "-p,    --prompt PROMPT",        "prompt to start generation with\n"
+                                                                        "in conversation mode, this will be used as system prompt\n"
+                                                                        "(default: '%s')", params.prompt.c_str() });
    options.push_back({ "*",           "-f,    --file FNAME",           "a file containing the prompt (default: none)" });
    options.push_back({ "*",           "       --in-file FNAME",        "an input file (repeat to specify multiple files)" });
    options.push_back({ "*",           "-bf,   --binary-file FNAME",    "binary file containing the prompt (default: none)" });
@@ -1402,13 +1435,17 @@ void gpt_params_print_usage(int /*argc*/, char ** argv, const gpt_params & param
                                                                        "halt generation at PROMPT, return control in interactive mode\n"
                                                                        "can be specified more than once for multiple prompts" });
    options.push_back({ "main",        "-sp,   --special",              "special tokens output enabled (default: %s)", params.special ? "true" : "false" });
-    options.push_back({ "main",        "-cnv,  --conversation",         "run in conversation mode (does not print special tokens and suffix/prefix) (default: %s)", params.conversation ? "true" : "false" });
+    options.push_back({ "main",        "-cnv,  --conversation",         "run in conversation mode, does not print special tokens and suffix/prefix\n"
+                                                                        "if suffix/prefix are not specified, default chat template will be used\n"
+                                                                        "(default: %s)", params.conversation ? "true" : "false" });
    options.push_back({ "main infill", "-i,    --interactive",          "run in interactive mode (default: %s)", params.interactive ? "true" : "false" });
    options.push_back({ "main infill", "-if,   --interactive-first",    "run in interactive mode and wait for input right away (default: %s)", params.interactive_first ? "true" : "false" });
    options.push_back({ "main infill", "-mli,  --multiline-input",      "allows you to write or paste multiple lines without ending each in '\\'" });
    options.push_back({ "main infill", "       --in-prefix-bos",        "prefix BOS to user inputs, preceding the `--in-prefix` string" });
    options.push_back({ "main infill", "       --in-prefix STRING",     "string to prefix user inputs with (default: empty)" });
    options.push_back({ "main infill", "       --in-suffix STRING",     "string to suffix after user inputs with (default: empty)" });
+    options.push_back({ "server infill",
+                                       "       --spm-infill",           "use Suffix/Prefix/Middle pattern for infill (instead of Prefix/Suffix/Middle) as some models prefer this. (default: %s)", params.spm_infill ? "enabled" : "disabled" });

    options.push_back({ "sampling" });
    options.push_back({ "*",           "       --samplers SAMPLERS",    "samplers that will be used for generation in the order, separated by \';\'\n"
@@ -1444,6 +1481,7 @@ void gpt_params_print_usage(int /*argc*/, char ** argv, const gpt_params & param
    options.push_back({ "main",        "       --cfg-scale N",          "strength of guidance (default: %.1f, 1.0 = disable)", (double)sparams.cfg_scale });
    options.push_back({ "main",        "       --chat-template JINJA_TEMPLATE",
                                                                        "set custom jinja chat template (default: template taken from model's metadata)\n"
+                                                                        "if suffix/prefix are specified, template will be disabled\n"
                                                                        "only commonly used templates are accepted:\n"
                                                                        "https://github.com/ggerganov/llama.cpp/wiki/Templates-supported-by-llama_chat_apply_template" });
    options.push_back({ "grammar" });
@@ -1454,8 +1492,10 @@ void gpt_params_print_usage(int /*argc*/, char ** argv, const gpt_params & param
                                                                        "For schemas w/ external $refs, use --grammar + example/json_schema_to_grammar.py instead" });

    options.push_back({ "embedding" });
-    options.push_back({ "embedding",   "       --pooling {none,mean,cls}",
+    options.push_back({ "embedding",   "       --pooling {none,mean,cls,last}",
                                                                        "pooling type for embeddings, use model default if unspecified" });
+    options.push_back({ "embedding",   "       --attention {causal,non-causal}",
+                                                                        "attention type for embeddings, use model default if unspecified" });

    options.push_back({ "context hacking" });
    options.push_back({ "*",           "       --rope-scaling {none,linear,yarn}",
@@ -1552,6 +1592,7 @@ void gpt_params_print_usage(int /*argc*/, char ** argv, const gpt_params & param
    options.push_back({ "*",           "-mu,   --model-url MODEL_URL",  "model download url (default: unused)" });
    options.push_back({ "*",           "-hfr,  --hf-repo REPO",         "Hugging Face model repository (default: unused)" });
    options.push_back({ "*",           "-hff,  --hf-file FILE",         "Hugging Face model file (default: unused)" });
+    options.push_back({ "*",           "-hft,  --hf-token TOKEN",       "Hugging Face access token (default: value from HF_TOKEN environment variable)" });

    options.push_back({ "retrieval" });
    options.push_back({ "retrieval",   "       --context-file FNAME",   "file to load context from (repeat to specify multiple files)" });
@@ -1991,9 +2032,9 @@ std::tuple<struct llama_model *, struct llama_context *> llama_init_from_gpt_par
    llama_model * model = nullptr;

    if (!params.hf_repo.empty() && !params.hf_file.empty()) {
-        model = llama_load_model_from_hf(params.hf_repo.c_str(), params.hf_file.c_str(), params.model.c_str(), mparams);
+        model = llama_load_model_from_hf(params.hf_repo.c_str(), params.hf_file.c_str(), params.model.c_str(), params.hf_token.c_str(), mparams);
    } else if (!params.model_url.empty()) {
-        model = llama_load_model_from_url(params.model_url.c_str(), params.model.c_str(), mparams);
+        model = llama_load_model_from_url(params.model_url.c_str(), params.model.c_str(), params.hf_token.c_str(), mparams);
    } else {
        model = llama_load_model_from_file(params.model.c_str(), mparams);
    }
@@ -2061,7 +2102,24 @@ std::tuple<struct llama_model *, struct llama_context *> llama_init_from_gpt_par
    if (params.warmup) {
        LOG("warming up the model with an empty run\n");

-        std::vector<llama_token> tmp = { llama_token_bos(model), llama_token_eos(model), };
+        std::vector<llama_token> tmp;
+        llama_token bos = llama_token_bos(model);
+        llama_token eos = llama_token_eos(model);
+        // some models (e.g. T5) don't have a BOS token
+        if (bos != -1) {
+            tmp.push_back(bos);
+        }
+        tmp.push_back(eos);
+
+        if (llama_model_has_encoder(model)) {
+            llama_encode(lctx, llama_batch_get_one(tmp.data(), tmp.size(), 0, 0));
+            llama_token decoder_start_token_id = llama_model_decoder_start_token(model);
+            if (decoder_start_token_id == -1) {
+                decoder_start_token_id = bos;
+            }
+            tmp.clear();
+            tmp.push_back(decoder_start_token_id);
+        }
        llama_decode(lctx, llama_batch_get_one(tmp.data(), std::min(tmp.size(), (size_t) params.n_batch), 0, 0));
        llama_kv_cache_clear(lctx);
        llama_synchronize(lctx);
@@ -2144,6 +2202,7 @@ struct llama_context_params llama_context_params_from_gpt_params(const gpt_param
    cparams.yarn_beta_slow    = params.yarn_beta_slow;
    cparams.yarn_orig_ctx     = params.yarn_orig_ctx;
    cparams.pooling_type      = params.pooling_type;
+    cparams.attention_type    = params.attention_type;
    cparams.defrag_thold      = params.defrag_thold;
    cparams.cb_eval           = params.cb_eval;
    cparams.cb_eval_user_data = params.cb_eval_user_data;
@@ -2163,7 +2222,7 @@ static bool starts_with(const std::string & str, const std::string & prefix) {
    return str.rfind(prefix, 0) == 0;
 }

-static bool llama_download_file(const std::string & url, const std::string & path) {
+static bool llama_download_file(const std::string & url, const std::string & path, const std::string & hf_token) {

    // Initialize libcurl
    std::unique_ptr<CURL, decltype(&curl_easy_cleanup)> curl(curl_easy_init(), &curl_easy_cleanup);
@@ -2178,6 +2237,15 @@ static bool llama_download_file(const std::string & url, const std::string & pat
    curl_easy_setopt(curl.get(), CURLOPT_URL, url.c_str());
    curl_easy_setopt(curl.get(), CURLOPT_FOLLOWLOCATION, 1L);

+    // Check if hf-token or bearer-token was specified
+    if (!hf_token.empty()) {
+      std::string auth_header = "Authorization: Bearer ";
+      auth_header += hf_token.c_str();
+      struct curl_slist *http_headers = NULL;
+      http_headers = curl_slist_append(http_headers, auth_header.c_str());
+      curl_easy_setopt(curl.get(), CURLOPT_HTTPHEADER, http_headers);
+    }
+
 #if defined(_WIN32)
    // CURLSSLOPT_NATIVE_CA tells libcurl to use standard certificate store of
    //   operating system. Currently implemented under MS-Windows.
@@ -2373,6 +2441,7 @@ static bool llama_download_file(const std::string & url, const std::string & pat
 struct llama_model * llama_load_model_from_url(
        const char * model_url,
        const char * path_model,
+        const char * hf_token,
        const struct llama_model_params & params) {
    // Basic validation of the model_url
    if (!model_url || strlen(model_url) == 0) {
@@ -2380,7 +2449,7 @@ struct llama_model * llama_load_model_from_url(
        return NULL;
    }

-    if (!llama_download_file(model_url, path_model)) {
+    if (!llama_download_file(model_url, path_model, hf_token)) {
        return NULL;
    }

@@ -2428,14 +2497,14 @@ struct llama_model * llama_load_model_from_url(
        // Prepare download in parallel
        std::vector<std::future<bool>> futures_download;
        for (int idx = 1; idx < n_split; idx++) {
-            futures_download.push_back(std::async(std::launch::async, [&split_prefix, &split_url_prefix, &n_split](int download_idx) -> bool {
+            futures_download.push_back(std::async(std::launch::async, [&split_prefix, &split_url_prefix, &n_split, hf_token](int download_idx) -> bool {
                char split_path[PATH_MAX] = {0};
                llama_split_path(split_path, sizeof(split_path), split_prefix, download_idx, n_split);

                char split_url[LLAMA_CURL_MAX_URL_LENGTH] = {0};
                llama_split_path(split_url, sizeof(split_url), split_url_prefix, download_idx, n_split);

-                return llama_download_file(split_url, split_path);
+                return llama_download_file(split_url, split_path, hf_token);
            }, idx));
        }

@@ -2454,6 +2523,7 @@ struct llama_model * llama_load_model_from_hf(
        const char * repo,
        const char * model,
        const char * path_model,
+        const char * hf_token,
        const struct llama_model_params & params) {
    // construct hugging face model url:
    //
@@ -2469,7 +2539,7 @@ struct llama_model * llama_load_model_from_hf(
    model_url += "/resolve/main/";
    model_url += model;

-    return llama_load_model_from_url(model_url.c_str(), path_model, params);
+    return llama_load_model_from_url(model_url.c_str(), path_model, hf_token, params);
 }

 #else
@@ -2477,6 +2547,7 @@ struct llama_model * llama_load_model_from_hf(
 struct llama_model * llama_load_model_from_url(
        const char * /*model_url*/,
        const char * /*path_model*/,
+        const char * /*hf_token*/,
        const struct llama_model_params & /*params*/) {
    fprintf(stderr, "%s: llama.cpp built without libcurl, downloading from an url not supported.\n", __func__);
    return nullptr;
@@ -2486,6 +2557,7 @@ struct llama_model * llama_load_model_from_hf(
        const char * /*repo*/,
        const char * /*model*/,
        const char * /*path_model*/,
+        const char * /*hf_token*/,
        const struct llama_model_params & /*params*/) {
    fprintf(stderr, "%s: llama.cpp built without libcurl, downloading from Hugging Face not supported.\n", __func__);
    return nullptr;
@@ -2550,51 +2622,35 @@ std::vector<llama_token> llama_tokenize(
 }

 std::string llama_token_to_piece(const struct llama_context * ctx, llama_token token, bool special) {
-    std::vector<char> result(8, 0);
-    const int n_tokens = llama_token_to_piece(llama_get_model(ctx), token, result.data(), result.size(), special);
-    if (n_tokens < 0) {
-        result.resize(-n_tokens);
-        int check = llama_token_to_piece(llama_get_model(ctx), token, result.data(), result.size(), special);
-        GGML_ASSERT(check == -n_tokens);
-    } else {
-        result.resize(n_tokens);
+    std::string piece;
+    piece.resize(piece.capacity());  // using string internal cache, 15 bytes + '\n'
+    const int n_chars = llama_token_to_piece(llama_get_model(ctx), token, &piece[0], piece.size(), 0, special);
+    if (n_chars < 0) {
+        piece.resize(-n_chars);
+        int check = llama_token_to_piece(llama_get_model(ctx), token, &piece[0], piece.size(), 0, special);
+        GGML_ASSERT(check == -n_chars);
+    }
+    else {
+        piece.resize(n_chars);
    }

-    return std::string(result.data(), result.size());
+    return piece;
 }

-std::string llama_detokenize_spm(llama_context * ctx, const std::vector<llama_token> & tokens) {
-    const llama_token bos_id = llama_token_bos(llama_get_model(ctx));
-
-    std::string piece;
-    std::string result;
-
-    for (size_t i = 0; i < tokens.size(); ++i) {
-        piece = llama_token_to_piece(ctx, tokens[i]);
-
-        // remove the leading space of the first non-BOS token
-        if (((tokens[0] == bos_id && i == 1) || (tokens[0] != bos_id && i == 0)) && piece[0] == ' ') {
-            piece = piece.substr(1);
-        }
-
-        result += piece;
+std::string llama_detokenize(llama_context * ctx, const std::vector<llama_token> & tokens, bool special) {
+    std::string text;
+    text.resize(std::max(text.capacity(), tokens.size()));
+    int32_t n_chars = llama_detokenize(llama_get_model(ctx), tokens.data(), (int32_t)tokens.size(), &text[0], (int32_t)text.size(), false, special);
+    if (n_chars < 0) {
+        text.resize(-n_chars);
+        n_chars = llama_detokenize(llama_get_model(ctx), tokens.data(), (int32_t)tokens.size(), &text[0], (int32_t)text.size(), false, special);
+        GGML_ASSERT(n_chars <= (int32_t)text.size());  // whitespace trimming is performed after per-token detokenization
    }

-    return result;
-}
-
-std::string llama_detokenize_bpe(llama_context * ctx, const std::vector<llama_token> & tokens) {
-    std::string piece;
-    std::string result;
-
-    for (size_t i = 0; i < tokens.size(); ++i) {
-        piece = llama_token_to_piece(ctx, tokens[i]);
-
-        result += piece;
-    }
+    text.resize(n_chars);

    // NOTE: the original tokenizer decodes bytes after collecting the pieces.
-    return result;
+    return text;
 }

 bool llama_should_add_bos_token(const llama_model * model) {
@@ -2618,6 +2674,7 @@ std::string llama_chat_apply_template(const struct llama_model * model,
        const std::vector<llama_chat_msg> & msgs,
        bool add_ass) {
    int alloc_size = 0;
+    bool fallback = false; // indicate if we must fallback to default chatml
    std::vector<llama_chat_message> chat;
    for (auto & msg : msgs) {
        chat.push_back({msg.role.c_str(), msg.content.c_str()});
@@ -2630,10 +2687,26 @@ std::string llama_chat_apply_template(const struct llama_model * model,
    // run the first time to get the total output length
    int32_t res = llama_chat_apply_template(model, ptr_tmpl, chat.data(), chat.size(), add_ass, buf.data(), buf.size());

+    // error: chat template is not supported
+    if (res < 0) {
+        if (ptr_tmpl != nullptr) {
+            // if the custom "tmpl" is not supported, we throw an error
+            // this is a bit redundant (for good), since we're not sure if user validated the custom template with llama_chat_verify_template()
+            throw std::runtime_error("this custom template is not supported");
+        } else {
+            // If the built-in template is not supported, we default to chatml
+            res = llama_chat_apply_template(nullptr, "chatml", chat.data(), chat.size(), add_ass, buf.data(), buf.size());
+            fallback = true;
+        }
+    }
+
    // if it turns out that our buffer is too small, we resize it
    if ((size_t) res > buf.size()) {
        buf.resize(res);
-        res = llama_chat_apply_template(model, ptr_tmpl, chat.data(), chat.size(), add_ass, buf.data(), buf.size());
+        res = llama_chat_apply_template(
+            fallback ? nullptr : model,
+            fallback ? "chatml" : ptr_tmpl,
+            chat.data(), chat.size(), add_ass, buf.data(), buf.size());
    }

    std::string formatted_chat(buf.data(), res);
@@ -2645,12 +2718,19 @@ std::string llama_chat_format_single(const struct llama_model * model,
        const std::vector<llama_chat_msg> & past_msg,
        const llama_chat_msg & new_msg,
        bool add_ass) {
+    std::ostringstream ss;
    auto fmt_past_msg = llama_chat_apply_template(model, tmpl, past_msg, false);
    std::vector<llama_chat_msg> chat_new(past_msg);
+    // if the past_msg ends with a newline, we must preserve it in the formatted version
+    if (add_ass && !fmt_past_msg.empty() && fmt_past_msg.back() == '\n') {
+        ss << "\n";
+    };
+    // format chat with new_msg
    chat_new.push_back(new_msg);
    auto fmt_new_msg = llama_chat_apply_template(model, tmpl, chat_new, add_ass);
-    auto formatted = fmt_new_msg.substr(fmt_past_msg.size(), fmt_new_msg.size() - fmt_past_msg.size());
-    return formatted;
+    // get the diff part
+    ss << fmt_new_msg.substr(fmt_past_msg.size(), fmt_new_msg.size() - fmt_past_msg.size());
+    return ss.str();
 }

 std::string llama_chat_format_example(const struct llama_model * model,
@@ -2804,125 +2884,87 @@ float llama_embd_similarity_cos(const float * embd1, const float * embd2, int n)
 //

 static llama_control_vector_data llama_control_vector_load_one(const llama_control_vector_load_info & load_info) {
-    int32_t n_tensors;
-
-    size_t n_bytes = 0;
-
-    uint32_t max_direction_layer = 0;
-
    llama_control_vector_data result = { -1, {} };

-    // calculate size of ctx needed for tensors, ensure tensors are f32, and find max layer
-    {
-        struct ggml_init_params meta_params = {
-            /* .mem_size   = */ ggml_tensor_overhead() * 128 + ggml_graph_overhead(),
-            /* .mem_buffer = */ nullptr,
-            /* .no_alloc   = */ true,
-        };
-        ggml_context * meta_ctx = ggml_init(meta_params);
-        struct gguf_init_params meta_gguf_params = {
-            /* .no_alloc = */ true,
-            /* .ctx      = */ &meta_ctx,
-        };
-        struct gguf_context * meta_ctx_gguf = gguf_init_from_file(load_info.fname.c_str(), meta_gguf_params);
-        if (!meta_ctx_gguf) {
-            fprintf(stderr, "%s: failed to load control vector from %s\n", __func__, load_info.fname.c_str());
-            ggml_free(meta_ctx);
-            return result;
-        }
-
-        n_tensors = gguf_get_n_tensors(meta_ctx_gguf);
-        for (int i = 0; i < n_tensors; i++) {
-            std::string name = gguf_get_tensor_name(meta_ctx_gguf, i);
-
-            // split on '.'
-            size_t dotpos = name.find('.');
-            if (dotpos != std::string::npos && name.substr(0, dotpos) == "direction") {
-                try {
-                    uint32_t layer = std::stoi(name.substr(dotpos + 1));
-                    if (layer == 0) {
-                        fprintf(stderr, "%s: direction tensor invalid in %s\n", __func__, load_info.fname.c_str());
-                        ggml_free(meta_ctx);
-                        gguf_free(meta_ctx_gguf);
-                        return result;
-                    }
-                    if (layer > max_direction_layer) {
-                        max_direction_layer = layer;
-                    }
-                } catch (...) {
-                    fprintf(stderr, "%s: direction tensor invalid in %s\n", __func__, load_info.fname.c_str());
-                    ggml_free(meta_ctx);
-                    gguf_free(meta_ctx_gguf);
-                    return result;
-                }
-            }
-
-            struct ggml_tensor * tensor_meta = ggml_get_tensor(meta_ctx, name.c_str());
-            if (tensor_meta->type != GGML_TYPE_F32 || ggml_n_dims(tensor_meta) != 1) {
-                fprintf(stderr, "%s: direction tensor invalid in %s\n", __func__, load_info.fname.c_str());
-                ggml_free(meta_ctx);
-                gguf_free(meta_ctx_gguf);
-                return result;
-            }
-            if (result.n_embd == -1) {
-                result.n_embd = ggml_nelements(tensor_meta);
-            } else if (ggml_nelements(tensor_meta) != result.n_embd) {
-                fprintf(stderr, "%s: direction tensor sizes mismatched in %s\n", __func__, load_info.fname.c_str());
-                ggml_free(meta_ctx);
-                gguf_free(meta_ctx_gguf);
-                return result;
-            }
-            n_bytes += ggml_nbytes(tensor_meta);
-        }
-        ggml_free(meta_ctx);
-        gguf_free(meta_ctx_gguf);
+    ggml_context * ctx = nullptr;
+    struct gguf_init_params meta_gguf_params = {
+        /* .no_alloc = */ false,
+        /* .ctx      = */ &ctx,
+    };
+    struct gguf_context * ctx_gguf = gguf_init_from_file(load_info.fname.c_str(), meta_gguf_params);
+    if (!ctx_gguf) {
+        fprintf(stderr, "%s: failed to load control vector file from %s\n", __func__, load_info.fname.c_str());
+        return result;
    }

+    int32_t n_tensors = gguf_get_n_tensors(ctx_gguf);
    if (n_tensors == 0) {
        fprintf(stderr, "%s: no direction tensors found in %s\n", __func__, load_info.fname.c_str());
-        return result;
    }

-    // load and scale tensors into final control vector context
-    struct ggml_init_params ggml_params = {
-        /* .mem_size   = */ ggml_tensor_overhead() * n_tensors + n_bytes,
-        /* .mem_buffer = */ nullptr,
-        /* .no_alloc   = */ false,
-    };
-    struct ggml_context * ctx = ggml_init(ggml_params);
+    for (int i = 0; i < n_tensors; i++) {
+        std::string name = gguf_get_tensor_name(ctx_gguf, i);

-    struct gguf_init_params params = {
-        /*.no_alloc = */ false,
-        /*.ctx      = */ &ctx,
-    };
-    struct gguf_context * ctx_gguf = gguf_init_from_file(load_info.fname.c_str(), params);
-    if (!ctx_gguf) {
-        fprintf(stderr, "%s: failed to load control vector from %s\n", __func__, load_info.fname.c_str());
-        ggml_free(ctx);
-        return result;
-    }
+        int layer_idx = -1;

-    // do not store data for layer 0 (it's not used)
-    result.data.resize(result.n_embd * max_direction_layer);
-
-    for (uint32_t il = 1; il <= max_direction_layer; il++) {
-        const std::string name = "direction." + std::to_string(il);
-        const ggml_tensor * tensor = ggml_get_tensor(ctx, name.c_str());
-
-        float * dst = result.data.data() + result.n_embd * (il - 1);
-
-        if (tensor) {
-            const float * src = (const float *) tensor->data;
-            for (int j = 0; j < result.n_embd; j++) {
-                dst[j] = src[j] * load_info.strength;
-            }
-        } else {
-            for (int j = 0; j < result.n_embd; j++) {
-                dst[j] = 0.0f;
+        // split on '.'
+        size_t dotpos = name.find('.');
+        if (dotpos != std::string::npos && name.substr(0, dotpos) == "direction") {
+            try {
+                layer_idx = std::stoi(name.substr(dotpos + 1));
+            } catch (...) {
+                layer_idx = -1;
            }
        }
+        if (layer_idx < 0) {
+            fprintf(stderr, "%s: invalid/unparsable direction tensor layer index in %s\n", __func__, load_info.fname.c_str());
+            result.n_embd = -1;
+            break;
+        } else if (layer_idx == 0) {
+            fprintf(stderr, "%s: invalid (zero) direction tensor layer index in %s\n", __func__, load_info.fname.c_str());
+            result.n_embd = -1;
+            break;
+        }
+
+        struct ggml_tensor * tensor = ggml_get_tensor(ctx, name.c_str());
+        if (tensor->type != GGML_TYPE_F32) {
+            fprintf(stderr, "%s: invalid (non-F32) direction tensor type in %s\n", __func__, load_info.fname.c_str());
+            result.n_embd = -1;
+            break;
+        }
+        if (ggml_n_dims(tensor) != 1) {
+            fprintf(stderr, "%s: invalid (non-1D) direction tensor shape in %s\n", __func__, load_info.fname.c_str());
+            result.n_embd = -1;
+            break;
+        }
+
+        if (result.n_embd == -1) {
+            result.n_embd = ggml_nelements(tensor);
+        } else if (ggml_nelements(tensor) != result.n_embd) {
+            fprintf(stderr, "%s: direction tensor in %s does not match previous dimensions\n", __func__, load_info.fname.c_str());
+            result.n_embd = -1;
+            break;
+        }
+
+        // extend if necessary - do not store data for layer 0 (it's not used)
+        result.data.resize(std::max(result.data.size(), static_cast<size_t>(result.n_embd * layer_idx)), 0.0f);
+
+        const float * src = (const float *) tensor->data;
+        float * dst = result.data.data() + result.n_embd * (layer_idx - 1);  // layer 1 at [0]
+        for (int j = 0; j < result.n_embd; j++) {
+            dst[j] += src[j] * load_info.strength;  // allows multiple directions for same layer in same file
+        }
+
    }

+    if (result.n_embd == -1) {
+        fprintf(stderr, "%s: skipping %s due to invalid direction tensors\n", __func__, load_info.fname.c_str());
+        result.data.clear();
+    }
+
+    gguf_free(ctx_gguf);
+    ggml_free(ctx);
+
    return result;
 }

@@ -2933,16 +2975,19 @@ llama_control_vector_data llama_control_vector_load(const std::vector<llama_cont
        auto cur = llama_control_vector_load_one(info);

        if (cur.n_embd == -1) {
-            return result;
+            result.n_embd = -1;
+            break;
        }
-        if (result.n_embd != -1 && (result.n_embd != cur.n_embd || result.data.size() != cur.data.size())) {
-            fprintf(stderr, "%s: control vector in %s does not match previous vector dimensions\n", __func__, info.fname.c_str());
-            return result;
+        if (result.n_embd != -1 && result.n_embd != cur.n_embd) {
+            fprintf(stderr, "%s: control vectors in %s does not match previous dimensions\n", __func__, info.fname.c_str());
+            result.n_embd = -1;
+            break;
        }

        if (result.n_embd == -1) {
            result = std::move(cur);
        } else {
+            result.data.resize(std::max(result.data.size(), cur.data.size()), 0.0f);  // extend if necessary
            for (size_t i = 0; i < cur.data.size(); i++) {
                result.data[i] += cur.data[i];
            }
@@ -2950,7 +2995,8 @@ llama_control_vector_data llama_control_vector_load(const std::vector<llama_cont
    }

    if (result.n_embd == -1) {
-        fprintf(stderr, "%s: no vectors passed\n", __func__);
+        fprintf(stderr, "%s: no valid control vector files passed\n", __func__);
+        result.data.clear();
    }

    return result;
@@ -99,6 +99,7 @@ struct gpt_params {
    enum llama_split_mode        split_mode        = LLAMA_SPLIT_MODE_LAYER; // how to split the model across GPUs
    enum llama_rope_scaling_type rope_scaling_type = LLAMA_ROPE_SCALING_TYPE_UNSPECIFIED;
    enum llama_pooling_type      pooling_type      = LLAMA_POOLING_TYPE_UNSPECIFIED; // pooling type for embeddings
+    enum llama_attention_type    attention_type    = LLAMA_ATTENTION_TYPE_UNSPECIFIED; // attention type for embeddings

    // // sampling parameters
    struct llama_sampling_params sparams;
@@ -107,6 +108,7 @@ struct gpt_params {
    std::string model_draft          = ""; // draft model for speculative decoding
    std::string model_alias          = "unknown"; // model alias
    std::string model_url            = ""; // model url to download
+    std::string hf_token             = ""; // HF token
    std::string hf_repo              = ""; // HF repo
    std::string hf_file              = ""; // HF file
    std::string prompt               = "";
@@ -200,6 +202,7 @@ struct gpt_params {
    std::string public_path   = "";
    std::string chat_template = "";
    std::string system_prompt = "";
+    bool enable_chat_template = true;

    std::vector<std::string> api_keys;

@@ -250,8 +253,11 @@ struct gpt_params {
    std::string cvector_outfile       = "control_vector.gguf";
    std::string cvector_positive_file = "examples/cvector-generator/positive.txt";
    std::string cvector_negative_file = "examples/cvector-generator/negative.txt";
+
+    bool spm_infill = false; // suffix/prefix/middle pattern for infill
 };

+void gpt_params_handle_hf_token(gpt_params & params);
 void gpt_params_handle_model_default(gpt_params & params);

 bool gpt_params_parse_ex   (int argc, char ** argv, gpt_params & params);
@@ -307,8 +313,8 @@ std::tuple<struct llama_model *, struct llama_context *> llama_init_from_gpt_par
 struct llama_model_params   llama_model_params_from_gpt_params  (const gpt_params & params);
 struct llama_context_params llama_context_params_from_gpt_params(const gpt_params & params);

-struct llama_model * llama_load_model_from_url(const char * model_url, const char * path_model, const struct llama_model_params & params);
-struct llama_model * llama_load_model_from_hf(const char * repo, const char * file, const char * path_model, const struct llama_model_params & params);
+struct llama_model * llama_load_model_from_url(const char * model_url, const char * path_model, const char * hf_token, const struct llama_model_params & params);
+struct llama_model * llama_load_model_from_hf(const char * repo, const char * file, const char * path_model, const char * hf_token, const struct llama_model_params & params);

 // Batch utils

@@ -346,21 +352,13 @@ std::string llama_token_to_piece(
                       llama_token   token,
                       bool          special = true);

-// TODO: these should be moved in llama.h C-style API under single `llama_detokenize` function
-//       that takes into account the tokenizer type and decides how to handle the leading space
-//
 // detokenizes a vector of tokens into a string
 // should work similar to Python's `tokenizer.decode`
-// removes the leading space from the first non-BOS token
-std::string llama_detokenize_spm(
+// optionally renders special/control tokens
+std::string llama_detokenize(
                         llama_context * ctx,
-        const std::vector<llama_token> & tokens);
-
-// detokenizes a vector of tokens into a string
-// should work similar to Python's `tokenizer.decode`
-std::string llama_detokenize_bpe(
-                         llama_context * ctx,
-        const std::vector<llama_token> & tokens);
+        const std::vector<llama_token> & tokens,
+                                  bool   special = true);

 // Uses the value from the model metadata if possible, otherwise
 // defaults to true when model type is SPM, otherwise false.
@@ -380,6 +378,8 @@ struct llama_chat_msg {
 bool llama_chat_verify_template(const std::string & tmpl);

 // CPP wrapper for llama_chat_apply_template
+// If the built-in template is not supported, we default to chatml
+// If the custom "tmpl" is not supported, we throw an error
 std::string llama_chat_apply_template(const struct llama_model * model,
        const std::string & tmpl,
        const std::vector<llama_chat_msg> & chat,
@@ -454,4 +454,3 @@ void yaml_dump_string_multiline(FILE * stream, const char * prop_name, const cha
 void yaml_dump_non_result_info(
    FILE * stream, const gpt_params & params, const llama_context * lctx,
    const std::string & timestamp, const std::vector<int> & prompt_tokens, const char * model_desc);
-
@@ -316,7 +316,7 @@ std::unordered_map<char, std::string> GRAMMAR_LITERAL_ESCAPES = {
 };

 std::unordered_set<char> NON_LITERAL_SET = {'|', '.', '(', ')', '[', ']', '{', '}', '*', '+', '?'};
-std::unordered_set<char> ESCAPED_IN_REGEXPS_BUT_NOT_IN_LITERALS = {'[', ']', '(', ')', '|', '{', '}', '*', '+', '?'};
+std::unordered_set<char> ESCAPED_IN_REGEXPS_BUT_NOT_IN_LITERALS = {'^', '$', '.', '[', ']', '(', ')', '|', '{', '}', '*', '+', '?'};

 template <typename Iterator>
 std::string join(Iterator begin, Iterator end, const std::string & separator) {
@@ -720,7 +720,7 @@ private:
            }
            prop_names.push_back(prop_name);
        }
-        if (!(additional_properties.is_boolean() && !additional_properties.get<bool>())) {
+        if ((additional_properties.is_boolean() && additional_properties.get<bool>()) || additional_properties.is_object()) {
            std::string sub_name = name + (name.empty() ? "" : "-") + "additional";
            std::string value_rule =
                additional_properties.is_object() ? visit(additional_properties, sub_name + "-value")
@@ -13,7 +13,7 @@ import sys
 from enum import IntEnum
 from pathlib import Path
 from hashlib import sha256
-from typing import TYPE_CHECKING, Any, Callable, ContextManager, Iterable, Iterator, Sequence, TypeVar, cast
+from typing import TYPE_CHECKING, Any, Callable, ContextManager, Iterable, Iterator, Literal, Sequence, TypeVar, cast

 import math
 import numpy as np
@@ -404,7 +404,7 @@ class Model:

        return tokens, toktypes, tokpre

-    # NOTE: this function is generated by convert-hf-to-gguf-update.py
+    # NOTE: this function is generated by convert_hf_to_gguf_update.py
    #       do not modify it manually!
    # ref:  https://github.com/ggerganov/llama.cpp/pull/6920
    # Marker: Start get_vocab_base_pre
@@ -424,7 +424,7 @@ class Model:

        res = None

-        # NOTE: if you get an error here, you need to update the convert-hf-to-gguf-update.py script
+        # NOTE: if you get an error here, you need to update the convert_hf_to_gguf_update.py script
        #       or pull the latest version of the model from Huggingface
        #       don't edit the hashes manually!
        if chkhsh == "0ef9807a4087ebef797fc749390439009c3b9eda9ad1a097abbe738f486c01e5":
@@ -487,15 +487,24 @@ class Model:
        if chkhsh == "7967bfa498ade6b757b064f31e964dddbb80f8f9a4d68d4ba7998fcf281c531a":
            # ref: https://huggingface.co/jinaai/jina-embeddings-v2-base-code
            res = "jina-v2-code"
+        if chkhsh == "b6e8e1518dc4305be2fe39c313ed643381c4da5db34a98f6a04c093f8afbe99b":
+            # ref: https://huggingface.co/THUDM/glm-4-9b-chat
+            res = "chatglm-bpe"
+        if chkhsh == "7fc505bd3104ca1083b150b17d088b59534ede9bde81f0dd2090967d7fe52cee":
+            # ref: https://huggingface.co/LumiOpen/Viking-7B
+            res = "viking"
+        if chkhsh == "b53802fb28e26d645c3a310b34bfe07da813026ec7c7716883404d5e0f8b1901":
+            # ref: https://huggingface.co/core42/jais-13b
+            res = "jais"

        if res is None:
            logger.warning("\n")
            logger.warning("**************************************************************************************")
            logger.warning("** WARNING: The BPE pre-tokenizer was not recognized!")
            logger.warning("**          There are 2 possible reasons for this:")
-            logger.warning("**          - the model has not been added to convert-hf-to-gguf-update.py yet")
+            logger.warning("**          - the model has not been added to convert_hf_to_gguf_update.py yet")
            logger.warning("**          - the pre-tokenization config has changed upstream")
-            logger.warning("**          Check your model files and convert-hf-to-gguf-update.py and update them accordingly.")
+            logger.warning("**          Check your model files and convert_hf_to_gguf_update.py and update them accordingly.")
            logger.warning("** ref:     https://github.com/ggerganov/llama.cpp/pull/6920")
            logger.warning("**")
            logger.warning(f"** chkhsh:  {chkhsh}")
@@ -573,7 +582,19 @@ class Model:
        special_vocab._set_special_token("unk", tokenizer.special_tokens["<|endoftext|>"])
        special_vocab.add_to_gguf(self.gguf_writer)

-    def _set_vocab_sentencepiece(self):
+    def _set_vocab_sentencepiece(self, add_to_gguf=True):
+        tokens, scores, toktypes = self._create_vocab_sentencepiece()
+
+        self.gguf_writer.add_tokenizer_model("llama")
+        self.gguf_writer.add_tokenizer_pre("default")
+        self.gguf_writer.add_token_list(tokens)
+        self.gguf_writer.add_token_scores(scores)
+        self.gguf_writer.add_token_types(toktypes)
+
+        special_vocab = gguf.SpecialVocab(self.dir_model, n_vocab=len(tokens))
+        special_vocab.add_to_gguf(self.gguf_writer)
+
+    def _create_vocab_sentencepiece(self):
        from sentencepiece import SentencePieceProcessor

        tokenizer_path = self.dir_model / 'tokenizer.model'
@@ -635,14 +656,7 @@ class Model:
                scores.append(-1000.0)
                toktypes.append(SentencePieceTokenTypes.UNUSED)

-        self.gguf_writer.add_tokenizer_model("llama")
-        self.gguf_writer.add_tokenizer_pre("default")
-        self.gguf_writer.add_token_list(tokens)
-        self.gguf_writer.add_token_scores(scores)
-        self.gguf_writer.add_token_types(toktypes)
-
-        special_vocab = gguf.SpecialVocab(self.dir_model, n_vocab=len(tokens))
-        special_vocab.add_to_gguf(self.gguf_writer)
+        return tokens, scores, toktypes

    def _set_vocab_llama_hf(self):
        vocab = gguf.LlamaHfVocab(self.dir_model)
@@ -666,6 +680,51 @@ class Model:
        special_vocab = gguf.SpecialVocab(self.dir_model, n_vocab=len(tokens))
        special_vocab.add_to_gguf(self.gguf_writer)

+    def _set_vocab_builtin(self, model_name: Literal["gpt-neox", "llama-spm"], vocab_size: int):
+        tokenizer_path = Path(sys.path[0]) / "models" / f"ggml-vocab-{model_name}.gguf"
+        logger.warning(f"Using tokenizer from '{os.path.relpath(tokenizer_path, os.getcwd())}'")
+        vocab_reader = gguf.GGUFReader(tokenizer_path, "r")
+
+        default_pre = "mpt" if model_name == "gpt-neox" else "default"
+
+        field = vocab_reader.get_field(gguf.Keys.Tokenizer.MODEL)
+        assert field  # tokenizer model
+        self.gguf_writer.add_tokenizer_model(bytes(field.parts[-1]).decode("utf-8"))
+
+        field = vocab_reader.get_field(gguf.Keys.Tokenizer.PRE)
+        self.gguf_writer.add_tokenizer_pre(bytes(field.parts[-1]).decode("utf-8") if field else default_pre)
+
+        field = vocab_reader.get_field(gguf.Keys.Tokenizer.LIST)
+        assert field  # token list
+        self.gguf_writer.add_token_list([bytes(field.parts[i]) for i in field.data][:vocab_size])
+
+        if model_name == "llama-spm":
+            field = vocab_reader.get_field(gguf.Keys.Tokenizer.SCORES)
+            assert field  # token scores
+            self.gguf_writer.add_token_scores([field.parts[i].tolist()[0] for i in field.data][:vocab_size])
+
+        field = vocab_reader.get_field(gguf.Keys.Tokenizer.TOKEN_TYPE)
+        assert field  # token types
+        self.gguf_writer.add_token_types([field.parts[i].tolist()[0] for i in field.data][:vocab_size])
+
+        if model_name != "llama-spm":
+            field = vocab_reader.get_field(gguf.Keys.Tokenizer.MERGES)
+            assert field  # token merges
+            self.gguf_writer.add_token_merges([bytes(field.parts[i]) for i in field.data])
+
+        if (field := vocab_reader.get_field(gguf.Keys.Tokenizer.BOS_ID)) is not None:
+            self.gguf_writer.add_bos_token_id(field.parts[-1].tolist()[0])
+        if (field := vocab_reader.get_field(gguf.Keys.Tokenizer.EOS_ID)) is not None:
+            self.gguf_writer.add_eos_token_id(field.parts[-1].tolist()[0])
+        if (field := vocab_reader.get_field(gguf.Keys.Tokenizer.UNK_ID)) is not None:
+            self.gguf_writer.add_unk_token_id(field.parts[-1].tolist()[0])
+        if (field := vocab_reader.get_field(gguf.Keys.Tokenizer.PAD_ID)) is not None:
+            self.gguf_writer.add_pad_token_id(field.parts[-1].tolist()[0])
+        if (field := vocab_reader.get_field(gguf.Keys.Tokenizer.ADD_BOS)) is not None:
+            self.gguf_writer.add_add_bos_token(field.parts[-1].tolist()[0])
+        if (field := vocab_reader.get_field(gguf.Keys.Tokenizer.ADD_EOS)) is not None:
+            self.gguf_writer.add_add_eos_token(field.parts[-1].tolist()[0])
+

@Model.register("GPTNeoXForCausalLM")
 class GPTNeoXModel(Model):
@@ -1931,7 +1990,7 @@ class Phi3MiniModel(Model):
        if len(rope_scaling_type) == 0:
            raise KeyError('Missing the required key rope_scaling.type')

-        if rope_scaling_type == 'su':
+        if rope_scaling_type == 'su' or rope_scaling_type == 'longrope':
            attn_factor = math.sqrt(1 + math.log(scale) / math.log(orig_max_pos_embds)) if scale > 1.0 else 1.0
        elif rope_scaling_type == 'yarn':
            attn_factor = 0.1 * math.log(scale) + 1.0 if scale > 1.0 else 1.0
@@ -2305,6 +2364,8 @@ class GemmaModel(Model):
        special_vocab._set_special_token("eot",    107)
        special_vocab.add_to_gguf(self.gguf_writer)

+        self.gguf_writer.add_add_space_prefix(False)
+
    def set_gguf_parameters(self):
        hparams = self.hparams
        block_count = hparams["num_hidden_layers"]
@@ -2337,6 +2398,71 @@ class GemmaModel(Model):
        return [(self.map_tensor_name(name), data_torch)]


+@Model.register("Gemma2ForCausalLM")
+class Gemma2Model(Model):
+    model_arch = gguf.MODEL_ARCH.GEMMA2
+
+    def set_vocab(self):
+        tokens, scores, toktypes = self._create_vocab_sentencepiece()
+        # hack: This is required so that we can properly use start/end-of-turn for chat template
+        for i in range(108):
+            # including <unusedX>, <start_of_turn>, <end_of_turn>
+            toktypes[i] = SentencePieceTokenTypes.CONTROL
+        self.gguf_writer.add_tokenizer_model("llama")
+        self.gguf_writer.add_tokenizer_pre("default")
+        self.gguf_writer.add_token_list(tokens)
+        self.gguf_writer.add_token_scores(scores)
+        self.gguf_writer.add_token_types(toktypes)
+
+        special_vocab = gguf.SpecialVocab(self.dir_model, n_vocab=len(tokens))
+        special_vocab.add_to_gguf(self.gguf_writer)
+
+        self.gguf_writer.add_add_space_prefix(False)
+
+    def set_gguf_parameters(self):
+        hparams = self.hparams
+        block_count = hparams["num_hidden_layers"]
+
+        self.gguf_writer.add_name(self.dir_model.name if self.model_name is None else self.model_name)
+        self.gguf_writer.add_context_length(hparams["max_position_embeddings"])
+        self.gguf_writer.add_embedding_length(hparams["hidden_size"])
+        self.gguf_writer.add_block_count(block_count)
+        self.gguf_writer.add_feed_forward_length(hparams["intermediate_size"])
+        self.gguf_writer.add_head_count(hparams["num_attention_heads"])
+        self.gguf_writer.add_head_count_kv(self.hparams["num_key_value_heads"] if "num_key_value_heads" in hparams else hparams["num_attention_heads"])
+        self.gguf_writer.add_layer_norm_rms_eps(self.hparams["rms_norm_eps"])
+        self.gguf_writer.add_key_length(hparams["head_dim"])
+        self.gguf_writer.add_value_length(hparams["head_dim"])
+        self.gguf_writer.add_file_type(self.ftype)
+        self.gguf_writer.add_attn_logit_softcapping(
+            self.hparams["attn_logit_softcapping"]
+        )
+        self.gguf_writer.add_final_logit_softcapping(
+            self.hparams["final_logit_softcapping"]
+        )
+        self.gguf_writer.add_sliding_window(self.hparams["sliding_window"])
+
+        # sanity check
+        attn_scalar = self.hparams["query_pre_attn_scalar"]
+        if attn_scalar != hparams["hidden_size"] / hparams["num_attention_heads"]:
+            raise ValueError("query_pre_attn_scalar must be equal to n_embd / n_head")
+
+    def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]:
+        del bid  # unused
+
+        # lm_head is not used in llama.cpp, while autoawq will include this tensor in model
+        # To prevent errors, skip loading lm_head.weight.
+        if name == "lm_head.weight":
+            logger.debug(f"Skipping get tensor {name!r} in safetensors so that convert can end normally.")
+            return []
+
+        # ref: https://github.com/huggingface/transformers/blob/fc37f38915372c15992b540dfcbbe00a916d4fc6/src/transformers/models/gemma/modeling_gemma.py#L89
+        if name.endswith("norm.weight"):
+            data_torch = data_torch + 1
+
+        return [(self.map_tensor_name(name), data_torch)]
+
+
@Model.register("Starcoder2ForCausalLM")
 class StarCoder2Model(Model):
    model_arch = gguf.MODEL_ARCH.STARCODER2
@@ -2361,39 +2487,7 @@ class MambaModel(Model):
            self._set_vocab_sentencepiece()
        else:
            # Use the GPT-NeoX tokenizer when no tokenizer files are present
-            tokenizer_path = Path(sys.path[0]) / "models" / "ggml-vocab-gpt-neox.gguf"
-            logger.warning(f"Using tokenizer from '{os.path.relpath(tokenizer_path, os.getcwd())}'")
-            neox_reader = gguf.GGUFReader(tokenizer_path, "r")
-
-            field = neox_reader.get_field(gguf.Keys.Tokenizer.MODEL)
-            self.gguf_writer.add_tokenizer_model(bytes(field.parts[-1]).decode("utf-8") if field else "gpt2")
-
-            field = neox_reader.get_field(gguf.Keys.Tokenizer.PRE)
-            self.gguf_writer.add_tokenizer_pre(bytes(field.parts[-1]).decode("utf-8") if field else "mpt")
-
-            field = neox_reader.get_field(gguf.Keys.Tokenizer.LIST)
-            assert field
-            self.gguf_writer.add_token_list([bytes(field.parts[i]) for i in field.data][:vocab_size])
-
-            field = neox_reader.get_field(gguf.Keys.Tokenizer.TOKEN_TYPE)
-            assert field
-            self.gguf_writer.add_token_types([field.parts[i].tolist()[0] for i in field.data][:vocab_size])
-
-            field = neox_reader.get_field(gguf.Keys.Tokenizer.MERGES)
-            assert field
-            self.gguf_writer.add_token_merges([bytes(field.parts[i]) for i in field.data])
-
-            field = neox_reader.get_field(gguf.Keys.Tokenizer.BOS_ID)
-            self.gguf_writer.add_bos_token_id(field.parts[-1].tolist()[0] if field else 1)
-
-            field = neox_reader.get_field(gguf.Keys.Tokenizer.EOS_ID)
-            self.gguf_writer.add_eos_token_id(field.parts[-1].tolist()[0] if field else 0)
-
-            field = neox_reader.get_field(gguf.Keys.Tokenizer.UNK_ID)
-            self.gguf_writer.add_unk_token_id(field.parts[-1].tolist()[0] if field else 0)
-
-            field = neox_reader.get_field(gguf.Keys.Tokenizer.PAD_ID)
-            self.gguf_writer.add_pad_token_id(field.parts[-1].tolist()[0] if field else 0)
+            self._set_vocab_builtin("gpt-neox", vocab_size)

    def set_gguf_parameters(self):
        d_model = self.find_hparam(["hidden_size",       "d_model"])
@@ -2545,6 +2639,82 @@ class JinaBertV2Model(BertModel):
        self.gguf_writer.add_add_eos_token(True)


+@Model.register("OpenELMForCausalLM")
+class OpenELMModel(Model):
+    model_arch = gguf.MODEL_ARCH.OPENELM
+
+    @staticmethod
+    def _make_divisible(v: float | int, divisor: int) -> int:
+        # ref: https://huggingface.co/apple/OpenELM-270M-Instruct/blob/eb111ff2e6724348e5b905984063d4064d4bc579/configuration_openelm.py#L34-L38
+        new_v = max(divisor, int(v + divisor / 2) // divisor * divisor)
+        # Make sure that round down does not go down by more than 10%.
+        if new_v < 0.9 * v:
+            new_v += divisor
+        return new_v
+
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+
+        ffn_multipliers: list[float] = self.hparams["ffn_multipliers"]
+        ffn_dim_divisor: int = self.hparams["ffn_dim_divisor"]
+        self._n_embd: int = self.hparams["model_dim"]
+        self._num_kv_heads: list[int] = self.hparams["num_kv_heads"]
+        self._num_query_heads: list[int] = self.hparams["num_query_heads"]
+        self._ffn_dims: list[int] = [
+            OpenELMModel._make_divisible(multiplier * self._n_embd, ffn_dim_divisor)
+            for multiplier in ffn_multipliers
+        ]
+        assert isinstance(self._num_kv_heads, list) and isinstance(self._num_kv_heads[0], int)
+        assert isinstance(self._num_query_heads, list) and isinstance(self._num_query_heads[0], int)
+
+    # Uses the tokenizer from meta-llama/Llama-2-7b-hf
+    def set_vocab(self):
+        try:
+            self._set_vocab_sentencepiece()
+        except FileNotFoundError:
+            self._set_vocab_builtin("llama-spm", self.hparams["vocab_size"])
+
+    def set_gguf_parameters(self):
+        n_embd = self._n_embd
+        head_dim = self.hparams["head_dim"]
+        rot_pct = 1.0
+        assert self.block_count == len(self._num_kv_heads)
+        assert self.block_count == len(self._num_query_heads)
+        assert self.block_count == len(self._ffn_dims)
+
+        self.gguf_writer.add_name(self.dir_model.name if self.model_name is None else self.model_name)
+        self.gguf_writer.add_block_count(self.block_count)
+        self.gguf_writer.add_context_length(self.hparams["max_context_length"])
+        self.gguf_writer.add_embedding_length(n_embd)
+        self.gguf_writer.add_feed_forward_length(self._ffn_dims)
+        self.gguf_writer.add_head_count(self._num_query_heads)
+        self.gguf_writer.add_head_count_kv(self._num_kv_heads)
+        self.gguf_writer.add_rope_freq_base(self.hparams["rope_freq_constant"])
+        # https://huggingface.co/apple/OpenELM-270M-Instruct/blob/c401df2/modeling_openelm.py#L30
+        self.gguf_writer.add_layer_norm_rms_eps(1e-6)
+        self.gguf_writer.add_rope_dimension_count(int(rot_pct * head_dim))
+        self.gguf_writer.add_key_length(head_dim)
+        self.gguf_writer.add_value_length(head_dim)
+        self.gguf_writer.add_file_type(self.ftype)
+
+    def find_hparam(self, keys: Iterable[str], optional: bool = False) -> Any:
+        if "n_layers" in keys:
+            return self.hparams["num_transformer_layers"]
+
+        return super().find_hparam(keys, optional)
+
+    def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]:
+
+        # split ff
+        if bid is not None and name == f"transformer.layers.{bid}.ffn.proj_1.weight":
+            ff_dim = self._ffn_dims[bid]
+            yield (self.format_tensor_name(gguf.MODEL_TENSOR.FFN_GATE, bid), data_torch[:ff_dim])
+            yield (self.format_tensor_name(gguf.MODEL_TENSOR.FFN_UP, bid), data_torch[ff_dim:])
+            return
+
+        yield (self.map_tensor_name(name), data_torch)
+
+
@Model.register("ArcticForCausalLM")
 class ArcticModel(Model):
    model_arch = gguf.MODEL_ARCH.ARCTIC
@@ -2775,11 +2945,17 @@ class DeepseekV2Model(Model):
                raise ValueError(f"Unprocessed experts: {experts}")


-@Model.register("T5ForConditionalGeneration")
@Model.register("T5WithLMHeadModel")
+@Model.register("T5ForConditionalGeneration")
+@Model.register("MT5ForConditionalGeneration")
+@Model.register("UMT5ForConditionalGeneration")
 class T5Model(Model):
    model_arch = gguf.MODEL_ARCH.T5

+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        self.shared_token_embeddings_found = False
+
    def set_vocab(self):
        # to avoid TypeError: Descriptors cannot be created directly
        # exception when importing sentencepiece_model_pb2
@@ -2787,17 +2963,29 @@ class T5Model(Model):
        from sentencepiece import SentencePieceProcessor
        from sentencepiece import sentencepiece_model_pb2 as model

-        tokenizer_path = self.dir_model / 'spiece.model'
+        tokenizer_path = self.dir_model / 'tokenizer.model'
+
+        # many older models use spiece.model tokenizer model filename
+        if not tokenizer_path.is_file():
+            tokenizer_path = self.dir_model / 'spiece.model'

        if not tokenizer_path.is_file():
            raise FileNotFoundError(f"File not found: {tokenizer_path}")

        sentencepiece_model = model.ModelProto()
        sentencepiece_model.ParseFromString(open(tokenizer_path, "rb").read())
+
+        # some models like Pile-T5 family use BPE tokenizer instead of Unigram
+        if sentencepiece_model.trainer_spec.model_type == 2: # BPE
+            # assure the tokenizer model file name is correct
+            assert tokenizer_path.name == 'tokenizer.model'
+            return self._set_vocab_sentencepiece()
+        else:
+            assert sentencepiece_model.trainer_spec.model_type == 1 # UNIGRAM
+
        add_prefix = sentencepiece_model.normalizer_spec.add_dummy_prefix
        remove_whitespaces = sentencepiece_model.normalizer_spec.remove_extra_whitespaces
        precompiled_charsmap = sentencepiece_model.normalizer_spec.precompiled_charsmap
-        assert sentencepiece_model.trainer_spec.model_type == 1 # UNIGRAM

        tokenizer = SentencePieceProcessor()
        tokenizer.LoadFromFile(str(tokenizer_path))
@@ -2867,7 +3055,10 @@ class T5Model(Model):

    def set_gguf_parameters(self):
        self.gguf_writer.add_name("T5")
-        self.gguf_writer.add_context_length(self.hparams["n_positions"])
+        if (n_ctx := self.find_hparam(["n_positions"], optional=True)) is None:
+            logger.warning("Couldn't find context length in config.json, assuming default value of 512")
+            n_ctx = 512
+        self.gguf_writer.add_context_length(n_ctx)
        self.gguf_writer.add_embedding_length(self.hparams["d_model"])
        self.gguf_writer.add_feed_forward_length(self.hparams["d_ff"])
        self.gguf_writer.add_block_count(self.hparams["num_layers"])
@@ -2883,16 +3074,295 @@ class T5Model(Model):
    def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]:
        del bid  # unused

-        # Sometimes T5 and Flan-T5 based models contain "encoder.embed_tokens.weight" tensor or
-        # "decoder.embed_tokens.weight" tensors that are duplicates of "shared.weight" tensor
-        # To prevent errors caused by an unnecessary unmapped tensor, skip both of them and use only "shared.weight".
-        if name == "decoder.embed_tokens.weight" or name == "encoder.embed_tokens.weight":
-            logger.debug(f"Skipping tensor {name!r} in safetensors so that convert can end normally.")
-            return []
+        # T5 based models contain shared token embeddings tensors saved randomly as either "encoder.embed_tokens.weight",
+        # "decoder.embed_tokens.weight" or "shared.weight" tensor. In some models there are even multiple of them stored
+        # in the safetensors files. We use the first tensor from these three as the token embeddings for both encoder
+        # and decoder and ignore the remaining ones.
+        if name in ["decoder.embed_tokens.weight", "encoder.embed_tokens.weight", "shared.weight"]:
+            if not self.shared_token_embeddings_found:
+                name = "shared.weight"
+                self.shared_token_embeddings_found = True
+            else:
+                logger.debug(f"Skipping shared tensor {name!r} in safetensors so that convert can end normally.")
+                return []

        return [(self.map_tensor_name(name), data_torch)]


+@Model.register("JAISLMHeadModel")
+class JaisModel(Model):
+    model_arch = gguf.MODEL_ARCH.JAIS
+
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+
+        # SwigLU activation
+        assert self.hparams["activation_function"] == "swiglu"
+        # ALiBi position embedding
+        assert self.hparams["position_embedding_type"] == "alibi"
+
+        # Embeddings scale
+        self.embeddings_scale = 1.0
+        # note: For some JAIS flavors, output is tied to (same as) wte in original model
+        self.output_is_wte = False
+        if 'mup_embeddings_scale' in self.hparams:
+            self.output_is_wte = True   # Hack (?)
+            self.embeddings_scale = self.hparams['mup_embeddings_scale']
+        elif 'embeddings_scale' in self.hparams:
+            self.embeddings_scale = self.hparams['embeddings_scale']
+        else:
+            assert False
+
+        self.width_scale = 1.0
+        if 'mup_output_alpha' in self.hparams:
+            assert 'mup_width_scale' in self.hparams
+            self.width_scale = self.hparams['mup_output_alpha'] * self.hparams['mup_width_scale']
+        elif 'width_scale' in self.hparams:
+            self.width_scale = self.hparams['width_scale']
+        else:
+            assert False
+
+        self.max_alibi_bias = 8.0
+
+    def set_vocab(self):
+        self._set_vocab_gpt2()
+
+    def set_gguf_parameters(self):
+        self.gguf_writer.add_name(self.dir_model.name)
+        self.gguf_writer.add_block_count(self.hparams["n_layer"])
+        self.gguf_writer.add_context_length(self.hparams["n_positions"])
+        self.gguf_writer.add_embedding_length(self.hparams["n_embd"])
+        self.gguf_writer.add_feed_forward_length(self.hparams["n_inner"])
+        self.gguf_writer.add_head_count(self.hparams["n_head"])
+        self.gguf_writer.add_layer_norm_eps(self.hparams["layer_norm_epsilon"])
+        self.gguf_writer.add_file_type(self.ftype)
+
+    def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]:
+        del bid  # unused
+
+        tensors: list[tuple[str, Tensor]] = []
+
+        # we don't need these
+        if name.endswith((".attn.bias")):
+            return tensors
+
+        if name.endswith(("relative_pe.slopes")):
+            # Calculate max ALiBi bias (this is the inverse of the ALiBi calculation)
+            # Some other models has max_alibi_bias spelled out explicitly in the hyperparams,
+            # but Jais's PyTorch model simply precalculates the slope values and places them
+            # in relative_pes.slopes
+            n_head_closest_log2 = 2 ** math.floor(math.log2(self.hparams["n_head"]))
+            first_val = float(data_torch._data[0])
+            self.max_alibi_bias = -round(math.log2(first_val) * n_head_closest_log2)
+
+            return tensors
+
+        if name.endswith((".c_attn.weight", ".c_proj.weight", ".c_fc.weight", ".c_fc2.weight")):
+            data_torch = data_torch.transpose(1, 0)
+
+        new_name = self.map_tensor_name(name)
+
+        if new_name == self.format_tensor_name(gguf.MODEL_TENSOR.TOKEN_EMBD):
+            tensors.append((new_name, data_torch * self.embeddings_scale))
+            if self.output_is_wte:
+                tensors.append((self.format_tensor_name(gguf.MODEL_TENSOR.OUTPUT), data_torch * self.width_scale))
+        elif new_name == self.format_tensor_name(gguf.MODEL_TENSOR.OUTPUT):
+            assert not self.output_is_wte
+            tensors.append((new_name, data_torch * self.width_scale))
+        else:
+            tensors.append((new_name, data_torch))
+
+        return tensors
+
+    def write_tensors(self):
+        super().write_tensors()
+        self.gguf_writer.add_max_alibi_bias(self.max_alibi_bias)
+
+
+@Model.register("ChatGLMModel", "ChatGLMForConditionalGeneration")
+class ChatGLMModel(Model):
+    model_arch = gguf.MODEL_ARCH.CHATGLM
+
+    def set_vocab_chatglm3(self):
+        dir_model = self.dir_model
+        hparams = self.hparams
+        tokens: list[bytearray] = []
+        toktypes: list[int] = []
+        scores: list[float] = []
+
+        from transformers import AutoTokenizer
+        tokenizer = AutoTokenizer.from_pretrained(dir_model, trust_remote_code=True)
+        vocab_size = hparams.get("padded_vocab_size", len(tokenizer.get_vocab()))
+        assert max(tokenizer.get_vocab().values()) < vocab_size
+        role_special_tokens = ["<|system|>", "<|user|>", "<|assistant|>", "<|observation|>"]
+        special_tokens = ["[MASK]", "[gMASK]", "[sMASK]", "sop", "eop"] + role_special_tokens
+        for token_id in range(vocab_size):
+            piece = tokenizer._convert_id_to_token(token_id)
+            if token_id == 0:
+                piece = "<unk>"
+            elif token_id == 1:
+                piece = "<bos>"
+            elif token_id == 2:
+                piece = "<eos>"
+
+            text = piece.encode("utf-8")
+            score = 0.0
+            # Referencing the tokenizer Python implementation(https://huggingface.co/THUDM/chatglm3-6b/blob/main/tokenization_chatglm.py),
+            # it is only valid if it is less than tokenizer.tokenizer.sp_model.vocab_size()
+            if len(piece) != 0 and token_id < tokenizer.tokenizer.sp_model.vocab_size():
+                score = tokenizer.tokenizer.sp_model.get_score(token_id)
+
+            if len(piece) == 0:
+                text = f"[PAD{token_id}]".encode("utf-8")
+
+            if token_id >= tokenizer.tokenizer.sp_model.vocab_size():
+                if piece in special_tokens:
+                    # show special tokens in prompt
+                    toktype = SentencePieceTokenTypes.USER_DEFINED
+                else:
+                    toktype = SentencePieceTokenTypes.UNKNOWN
+                tokens.append(text)
+                scores.append(score)
+                toktypes.append(toktype)
+                continue
+
+            toktype = SentencePieceTokenTypes.NORMAL
+            if tokenizer.tokenizer.sp_model.is_unknown(token_id):
+                toktype = SentencePieceTokenTypes.UNKNOWN
+            elif tokenizer.tokenizer.sp_model.is_control(token_id):
+                toktype = SentencePieceTokenTypes.CONTROL
+            elif tokenizer.tokenizer.sp_model.is_unused(token_id):
+                toktype = SentencePieceTokenTypes.UNUSED
+            elif tokenizer.tokenizer.sp_model.is_byte(token_id):
+                toktype = SentencePieceTokenTypes.BYTE
+
+            tokens.append(text)
+            scores.append(score)
+            toktypes.append(toktype)
+
+        self.gguf_writer.add_tokenizer_model("llama")
+        # glm3 needs prefix and suffix formatted as:
+        # prompt = "[gMASK]sop<|user|>\n" + prompt + "<|assistant|>"
+        self.gguf_writer.add_tokenizer_pre("chatglm-spm")
+        self.gguf_writer.add_token_list(tokens)
+        self.gguf_writer.add_token_scores(scores)
+        self.gguf_writer.add_token_types(toktypes)
+
+        special_vocab = gguf.SpecialVocab(self.dir_model, n_vocab=len(tokens))
+        special_vocab.add_to_gguf(self.gguf_writer)
+
+    @staticmethod
+    def token_bytes_to_string(b):
+        from transformers.models.gpt2.tokenization_gpt2 import bytes_to_unicode
+        byte_encoder = bytes_to_unicode()
+        return ''.join([byte_encoder[ord(char)] for char in b.decode('latin-1')])
+
+    @staticmethod
+    def bpe(mergeable_ranks: dict[bytes, int], token: bytes, max_rank: int | None = None) -> list[bytes]:
+        parts = [bytes([b]) for b in token]
+        while True:
+            min_idx = None
+            min_rank = None
+            for i, pair in enumerate(zip(parts[:-1], parts[1:])):
+                rank = mergeable_ranks.get(pair[0] + pair[1])
+                if rank is not None and (min_rank is None or rank < min_rank):
+                    min_idx = i
+                    min_rank = rank
+            if min_rank is None or (max_rank is not None and min_rank >= max_rank):
+                break
+            assert min_idx is not None
+            parts = parts[:min_idx] + [parts[min_idx] + parts[min_idx + 1]] + parts[min_idx + 2:]
+        return parts
+
+    def set_vocab(self):
+        if "THUDM/chatglm3-6b" in self.hparams.get("_name_or_path", ""):
+            self.set_vocab_chatglm3()
+            return
+
+        dir_model = self.dir_model
+        hparams = self.hparams
+        tokens: list[str] = []
+        toktypes: list[int] = []
+
+        from transformers import AutoTokenizer
+        tokenizer = AutoTokenizer.from_pretrained(dir_model, trust_remote_code=True)
+        vocab_size = hparams["padded_vocab_size"]
+        assert max(tokenizer.get_vocab().values()) < vocab_size
+
+        tokpre = self.get_vocab_base_pre(tokenizer)
+
+        merges = []
+        vocab = {}
+        mergeable_ranks = tokenizer.mergeable_ranks
+        for token, rank in mergeable_ranks.items():
+            vocab[ChatGLMModel.token_bytes_to_string(token)] = rank
+            if len(token) == 1:
+                continue
+            merged = ChatGLMModel.bpe(mergeable_ranks, token, max_rank=rank)
+            assert len(merged) >= 2 and len(merged) <= 7
+            merges.append(' '.join(map(ChatGLMModel.token_bytes_to_string, merged)))
+
+        # for this kind of tokenizer, added_vocab is not a subset of vocab, so they need to be combined
+        added_vocab = tokenizer.get_added_vocab()
+        reverse_vocab = {id_ : encoded_tok for encoded_tok, id_ in {**vocab, **added_vocab}.items()}
+
+        for i in range(vocab_size):
+            if i not in reverse_vocab:
+                tokens.append(f"[PAD{i}]")
+                toktypes.append(gguf.TokenType.USER_DEFINED)
+            elif reverse_vocab[i] in added_vocab:
+                tokens.append(reverse_vocab[i])
+                if tokenizer.added_tokens_decoder[i].special:
+                    toktypes.append(gguf.TokenType.CONTROL)
+                else:
+                    toktypes.append(gguf.TokenType.USER_DEFINED)
+            else:
+                tokens.append(reverse_vocab[i])
+                toktypes.append(gguf.TokenType.NORMAL)
+
+        self.gguf_writer.add_tokenizer_model("gpt2")
+        self.gguf_writer.add_tokenizer_pre(tokpre)
+        self.gguf_writer.add_token_list(tokens)
+        self.gguf_writer.add_token_types(toktypes)
+
+        special_vocab = gguf.SpecialVocab(dir_model, load_merges=False)
+        special_vocab.merges = merges
+        # only add special tokens when they were not already loaded from config.json
+        special_vocab._set_special_token("eos", tokenizer.get_added_vocab()["<|endoftext|>"])
+        special_vocab._set_special_token("eot", tokenizer.get_added_vocab()["<|user|>"])
+        # this one is usually not in config.json anyway
+        special_vocab._set_special_token("unk", tokenizer.get_added_vocab()["<|endoftext|>"])
+        special_vocab.add_to_gguf(self.gguf_writer)
+
+    def set_gguf_parameters(self):
+        self.gguf_writer.add_name(self.hparams.get("_name_or_path").split("/")[1]) # THUDM/glm4-9b-chat or THUDM/chatglm3-6b
+        n_embed = self.hparams.get("hidden_size", self.hparams.get("n_embed"))
+        n_head = self.hparams.get("n_head", self.hparams.get("num_attention_heads"))
+        n_head_kv = self.hparams.get("multi_query_group_num", n_head)
+        self.gguf_writer.add_context_length(self.hparams.get("seq_length", n_embed))
+        self.gguf_writer.add_embedding_length(n_embed)
+        self.gguf_writer.add_feed_forward_length(self.hparams.get("ffn_hidden_size", 4 * n_embed))
+        self.gguf_writer.add_block_count(self.hparams["num_layers"])
+        self.gguf_writer.add_head_count(n_head)
+        self.gguf_writer.add_head_count_kv(n_head_kv)
+        self.gguf_writer.add_layer_norm_rms_eps(self.hparams["layernorm_epsilon"])
+        self.gguf_writer.add_file_type(self.ftype)
+        self.gguf_writer.add_rope_dimension_count(64)
+        self.gguf_writer.add_add_bos_token(False)
+        rope_freq = 10000
+        if "rope_ratio" in self.hparams:
+            rope_freq = rope_freq * self.hparams["rope_ratio"]
+        self.gguf_writer.add_rope_freq_base(rope_freq)
+
+    def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]:
+        del bid  # unused
+
+        if name.endswith(".rotary_pos_emb.inv_freq"):
+            return []
+
+        name = name.removeprefix("transformer.")
+        return [(self.map_tensor_name(name), data_torch)]
+
 ###### CONVERSION LOGIC ######


@@ -2942,10 +3412,6 @@ def parse_args() -> argparse.Namespace:
        "--vocab-only", action="store_true",
        help="extract only the vocab",
    )
-    parser.add_argument(
-        "--awq-path", type=Path, default=None,
-        help="Path to scale awq cache file",
-    )
    parser.add_argument(
        "--outfile", type=Path,
        help="path to write to; default: based on input. {ftype} will be replaced by the outtype.",
@@ -3023,19 +3489,6 @@ def main() -> None:

    dir_model = args.model

-    if args.awq_path:
-        sys.path.insert(1, str(Path(__file__).parent / 'awq-py'))
-        from awq.apply_awq import add_scale_weights  # type: ignore[import-not-found]
-        tmp_model_path = args.model / "weighted_model"
-        dir_model = tmp_model_path
-        if tmp_model_path.is_dir():
-            logger.info(f"{tmp_model_path} exists as a weighted model.")
-        else:
-            tmp_model_path.mkdir(parents=True, exist_ok=True)
-            logger.info("Saving new weighted model ...")
-            add_scale_weights(str(args.model), str(args.awq_path), str(tmp_model_path))
-            logger.info(f"Saved weighted model at {tmp_model_path}.")
-
    if not dir_model.is_dir():
        logger.error(f'Error: {args.model} is not a directory')
        sys.exit(1)
@@ -3048,7 +3501,8 @@ def main() -> None:
        "auto": gguf.LlamaFileType.GUESSED,
    }

-    if args.use_temp_file and (args.split_max_tensors > 0 or args.split_max_size != "0"):
+    is_split = args.split_max_tensors > 0 or args.split_max_size != "0"
+    if args.use_temp_file and is_split:
        logger.error("Error: Cannot use temp file when splitting")
        sys.exit(1)

@@ -3085,11 +3539,12 @@ def main() -> None:
        if args.vocab_only:
            logger.info("Exporting model vocab...")
            model_instance.write_vocab()
-            logger.info("Model vocab successfully exported.")
+            logger.info(f"Model vocab successfully exported to {model_instance.fname_out}")
        else:
            logger.info("Exporting model...")
            model_instance.write()
-            logger.info("Model successfully exported.")
+            out_path = f"{model_instance.fname_out.parent}{os.sep}" if is_split else model_instance.fname_out
+            logger.info(f"Model successfully exported to {out_path}")


 if __name__ == '__main__':
@@ -2,7 +2,7 @@
 # -*- coding: utf-8 -*-

 # This script downloads the tokenizer models of the specified models from Huggingface and
-# generates the get_vocab_base_pre() function for convert-hf-to-gguf.py
+# generates the get_vocab_base_pre() function for convert_hf_to_gguf.py
 #
 # This is necessary in order to analyze the type of pre-tokenizer used by the model and
 # provide the necessary information to llama.cpp via the GGUF header in order to implement
@@ -15,9 +15,9 @@
 # - Add a new model to the "models" list
 # - Run the script with your huggingface token:
 #
-#   python3 convert-hf-to-gguf-update.py <huggingface_token>
+#   python3 convert_hf_to_gguf_update.py <huggingface_token>
 #
-# - Copy-paste the generated get_vocab_base_pre() function into convert-hf-to-gguf.py
+# - Copy-paste the generated get_vocab_base_pre() function into convert_hf_to_gguf.py
 # - Update llama.cpp with the new pre-tokenizer if necessary
 #
 # TODO: generate tokenizer tests for llama.cpp
@@ -37,7 +37,7 @@ from enum import IntEnum, auto
 from transformers import AutoTokenizer

 logging.basicConfig(level=logging.DEBUG)
-logger = logging.getLogger("convert-hf-to-gguf-update")
+logger = logging.getLogger("convert_hf_to_gguf_update")
 sess = requests.Session()


@@ -45,6 +45,7 @@ class TOKENIZER_TYPE(IntEnum):
    SPM = auto()
    BPE = auto()
    WPM = auto()
+    UGM = auto()


 # TODO: this string has to exercise as much pre-tokenizer functionality as possible
@@ -55,10 +56,10 @@ if len(sys.argv) == 2:
    token = sys.argv[1]
    if not token.startswith("hf_"):
        logger.info("Huggingface token seems invalid")
-        logger.info("Usage: python convert-hf-to-gguf-update.py <huggingface_token>")
+        logger.info("Usage: python convert_hf_to_gguf_update.py <huggingface_token>")
        sys.exit(1)
 else:
-    logger.info("Usage: python convert-hf-to-gguf-update.py <huggingface_token>")
+    logger.info("Usage: python convert_hf_to_gguf_update.py <huggingface_token>")
    sys.exit(1)

 # TODO: add models here, base models preferred
@@ -85,6 +86,11 @@ models = [
    {"name": "smaug-bpe",      "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/abacusai/Smaug-Llama-3-70B-Instruct", },
    {"name": "poro-chat",      "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/LumiOpen/Poro-34B-chat", },
    {"name": "jina-v2-code",   "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/jinaai/jina-embeddings-v2-base-code", },
+    {"name": "viking",         "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/LumiOpen/Viking-7B", }, # Also used for Viking 13B and 33B
+    {"name": "gemma",          "tokt": TOKENIZER_TYPE.SPM, "repo": "https://huggingface.co/google/gemma-2b", },
+    {"name": "gemma-2",        "tokt": TOKENIZER_TYPE.SPM, "repo": "https://huggingface.co/google/gemma-2-9b", },
+    {"name": "jais",           "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/core42/jais-13b", },
+    {"name": "t5",             "tokt": TOKENIZER_TYPE.UGM, "repo": "https://huggingface.co/google-t5/t5-small", },
 ]


@@ -106,9 +112,13 @@ def download_model(model):
    os.makedirs(f"models/tokenizers/{name}", exist_ok=True)

    files = ["config.json", "tokenizer.json", "tokenizer_config.json"]
+
    if tokt == TOKENIZER_TYPE.SPM:
        files.append("tokenizer.model")

+    if tokt == TOKENIZER_TYPE.UGM:
+        files.append("spiece.model")
+
    for file in files:
        save_path = f"models/tokenizers/{name}/{file}"
        if os.path.isfile(save_path):
@@ -124,14 +134,14 @@ for model in models:
        logger.error(f"Failed to download model {model['name']}. Error: {e}")


-# generate the source code for the convert-hf-to-gguf.py:get_vocab_base_pre() function:
+# generate the source code for the convert_hf_to_gguf.py:get_vocab_base_pre() function:

 src_ifs = ""
 for model in models:
    name = model["name"]
    tokt = model["tokt"]

-    if tokt == TOKENIZER_TYPE.SPM:
+    if tokt == TOKENIZER_TYPE.SPM or tokt == TOKENIZER_TYPE.UGM:
        continue

    # Skip if the tokenizer folder does not exist or there are other download issues previously
@@ -141,7 +151,10 @@ for model in models:

    # create the tokenizer
    try:
-        tokenizer = AutoTokenizer.from_pretrained(f"models/tokenizers/{name}")
+        if name == "t5":
+            tokenizer = AutoTokenizer.from_pretrained(f"models/tokenizers/{name}", use_fast=False)
+        else:
+            tokenizer = AutoTokenizer.from_pretrained(f"models/tokenizers/{name}")
    except OSError as e:
        logger.error(f"Error loading tokenizer for model {name}. The model may not exist or is not accessible with the provided token. Error: {e}")
        continue  # Skip to the next model if the tokenizer can't be loaded
@@ -188,7 +201,7 @@ src_func = f"""

        res = None

-        # NOTE: if you get an error here, you need to update the convert-hf-to-gguf-update.py script
+        # NOTE: if you get an error here, you need to update the convert_hf_to_gguf_update.py script
        #       or pull the latest version of the model from Huggingface
        #       don't edit the hashes manually!
 {src_ifs}
@@ -197,9 +210,9 @@ src_func = f"""
            logger.warning("**************************************************************************************")
            logger.warning("** WARNING: The BPE pre-tokenizer was not recognized!")
            logger.warning("**          There are 2 possible reasons for this:")
-            logger.warning("**          - the model has not been added to convert-hf-to-gguf-update.py yet")
+            logger.warning("**          - the model has not been added to convert_hf_to_gguf_update.py yet")
            logger.warning("**          - the pre-tokenization config has changed upstream")
-            logger.warning("**          Check your model files and convert-hf-to-gguf-update.py and update them accordingly.")
+            logger.warning("**          Check your model files and convert_hf_to_gguf_update.py and update them accordingly.")
            logger.warning("** ref:     https://github.com/ggerganov/llama.cpp/pull/6920")
            logger.warning("**")
            logger.warning(f"** chkhsh:  {{chkhsh}}")
@@ -213,7 +226,7 @@ src_func = f"""
        return res
 """

-convert_py_pth = pathlib.Path("convert-hf-to-gguf.py")
+convert_py_pth = pathlib.Path("convert_hf_to_gguf.py")
 convert_py = convert_py_pth.read_text(encoding="utf-8")
 convert_py = re.sub(
    r"(# Marker: Start get_vocab_base_pre)(.+?)( +# Marker: End get_vocab_base_pre)",
@@ -224,7 +237,7 @@ convert_py = re.sub(

 convert_py_pth.write_text(convert_py, encoding="utf-8")

-logger.info("+++ convert-hf-to-gguf.py was updated")
+logger.info("+++ convert_hf_to_gguf.py was updated")

 # generate tests for each tokenizer model

@@ -262,6 +275,7 @@ tests = [
    "\n =",
    "' era",
    "Hello, y'all! How are you 😁 ?我想在apple工作1314151天～",
+    "!!!!!!",
    "3",
    "33",
    "333",
@@ -271,7 +285,8 @@ tests = [
    "3333333",
    "33333333",
    "333333333",
-    # "Cửa Việt", # llama-bpe fails on this
+    "Cửa Việt", # llama-bpe fails on this
+    " discards",
    chktxt,
 ]

@@ -299,7 +314,10 @@ for model in models:

    # create the tokenizer
    try:
-        tokenizer = AutoTokenizer.from_pretrained(f"models/tokenizers/{name}")
+        if name == "t5":
+            tokenizer = AutoTokenizer.from_pretrained(f"models/tokenizers/{name}", use_fast=False)
+        else:
+            tokenizer = AutoTokenizer.from_pretrained(f"models/tokenizers/{name}")
    except OSError as e:
        logger.error(f"Failed to load tokenizer for model {name}. Error: {e}")
        continue  # Skip this model and continue with the next one in the loop
@@ -325,6 +343,6 @@ logger.info("\nRun the following commands to generate the vocab files for testin
 for model in models:
    name = model["name"]

-    print(f"python3 convert-hf-to-gguf.py models/tokenizers/{name}/ --outfile models/ggml-vocab-{name}.gguf --vocab-only") # noqa: NP100
+    print(f"python3 convert_hf_to_gguf.py models/tokenizers/{name}/ --outfile models/ggml-vocab-{name}.gguf --vocab-only") # noqa: NP100

 logger.info("\n")
@@ -0,0 +1,56 @@
+
+# Android
+
+## Build on Android using Termux
+[Termux](https://github.com/termux/termux-app#installation) is a method to execute `llama.cpp` on an Android device (no root required).
+```
+apt update && apt upgrade -y
+apt install git make cmake
+```
+
+It's recommended to move your model inside the `~/` directory for best performance:
+```
+cd storage/downloads
+mv model.gguf ~/
+```
+
+[Get the code](https://github.com/ggerganov/llama.cpp#get-the-code) & [follow the Linux build instructions](https://github.com/ggerganov/llama.cpp#build) to build `llama.cpp`.
+
+## Building the Project using Android NDK
+Obtain the [Android NDK](https://developer.android.com/ndk) and then build with CMake.
+
+Execute the following commands on your computer to avoid downloading the NDK to your mobile. Alternatively, you can also do this in Termux:
+```
+$ mkdir build-android
+$ cd build-android
+$ export NDK=<your_ndk_directory>
+$ cmake -DCMAKE_TOOLCHAIN_FILE=$NDK/build/cmake/android.toolchain.cmake -DANDROID_ABI=arm64-v8a -DANDROID_PLATFORM=android-23 -DCMAKE_C_FLAGS=-march=armv8.4a+dotprod ..
+$ make
+```
+
+Install [termux](https://github.com/termux/termux-app#installation) on your device and run `termux-setup-storage` to get access to your SD card (if Android 11+ then run the command twice).
+
+Finally, copy these built `llama` binaries and the model file to your device storage. Because the file permissions in the Android sdcard cannot be changed, you can copy the executable files to the `/data/data/com.termux/files/home/bin` path, and then execute the following commands in Termux to add executable permission:
+
+(Assumed that you have pushed the built executable files to the /sdcard/llama.cpp/bin path using `adb push`)
+```
+$cp -r /sdcard/llama.cpp/bin /data/data/com.termux/files/home/
+$cd /data/data/com.termux/files/home/bin
+$chmod +x ./*
+```
+
+Download model [llama-2-7b-chat.Q4_K_M.gguf](https://huggingface.co/TheBloke/Llama-2-7B-Chat-GGUF/blob/main/llama-2-7b-chat.Q4_K_M.gguf), and push it to `/sdcard/llama.cpp/`, then move it to `/data/data/com.termux/files/home/model/`
+
+```
+$mv /sdcard/llama.cpp/llama-2-7b-chat.Q4_K_M.gguf /data/data/com.termux/files/home/model/
+```
+
+Now, you can start chatting:
+```
+$cd /data/data/com.termux/files/home/bin
+$./llama-cli -m ../model/llama-2-7b-chat.Q4_K_M.gguf -n 128 -cml
+```
+
+Here's a demo of an interactive session running on Pixel 5 phone:
+
+https://user-images.githubusercontent.com/271616/225014776-1d567049-ad71-4ef2-b050-55b0b3b9274c.mp4
@@ -0,0 +1,288 @@
+# Build llama.cpp locally
+
+**To get the Code:**
+
+```bash
+git clone https://github.com/ggerganov/llama.cpp
+cd llama.cpp
+```
+
+In order to build llama.cpp you have four different options.
+
+- Using `make`:
+  - On Linux or MacOS:
+
+      ```bash
+      make
+      ```
+
+  - On Windows:
+
+    1. Download the latest fortran version of [w64devkit](https://github.com/skeeto/w64devkit/releases).
+    2. Extract `w64devkit` on your pc.
+    3. Run `w64devkit.exe`.
+    4. Use the `cd` command to reach the `llama.cpp` folder.
+    5. From here you can run:
+        ```bash
+        make
+        ```
+
+  - Notes:
+    - For faster compilation, add the `-j` argument to run multiple jobs in parallel. For example, `make -j 8` will run 8 jobs in parallel.
+    - For faster repeated compilation, install [ccache](https://ccache.dev/).
+    - For debug builds, run `make LLAMA_DEBUG=1`
+
+- Using `CMake`:
+
+  ```bash
+  cmake -B build
+  cmake --build build --config Release
+  ```
+
+  **Notes**:
+
+    - For faster compilation, add the `-j` argument to run multiple jobs in parallel. For example, `cmake --build build --config Release -j 8` will run 8 jobs in parallel.
+    - For faster repeated compilation, install [ccache](https://ccache.dev/).
+    - For debug builds, there are two cases:
+
+      1. Single-config generators (e.g. default = `Unix Makefiles`; note that they just ignore the `--config` flag):
+
+      ```bash
+      cmake -B build -DCMAKE_BUILD_TYPE=Debug
+      cmake --build build
+      ```
+
+      2. Multi-config generators (`-G` param set to Visual Studio, XCode...):
+
+      ```bash
+      cmake -B build -G "Xcode"
+      cmake --build build --config Debug
+      ```
+
+-   Using `gmake` (FreeBSD):
+
+    1. Install and activate [DRM in FreeBSD](https://wiki.freebsd.org/Graphics)
+    2. Add your user to **video** group
+    3. Install compilation dependencies.
+
+        ```bash
+        sudo pkg install gmake automake autoconf pkgconf llvm15 openblas
+
+        gmake CC=/usr/local/bin/clang15 CXX=/usr/local/bin/clang++15 -j4
+        ```
+
+## Metal Build
+
+On MacOS, Metal is enabled by default. Using Metal makes the computation run on the GPU.
+To disable the Metal build at compile time use the `GGML_NO_METAL=1` flag or the `GGML_METAL=OFF` cmake option.
+
+When built with Metal support, you can explicitly disable GPU inference with the `--n-gpu-layers|-ngl 0` command-line
+argument.
+
+## BLAS Build
+
+Building the program with BLAS support may lead to some performance improvements in prompt processing using batch sizes higher than 32 (the default is 512). Support with CPU-only BLAS implementations doesn't affect the normal generation performance. We may see generation performance improvements with GPU-involved BLAS implementations, e.g. cuBLAS, hipBLAS. There are currently several different BLAS implementations available for build and use:
+
+### Accelerate Framework:
+
+This is only available on Mac PCs and it's enabled by default. You can just build using the normal instructions.
+
+### OpenBLAS:
+
+This provides BLAS acceleration using only the CPU. Make sure to have OpenBLAS installed on your machine.
+
+- Using `make`:
+  - On Linux:
+    ```bash
+    make GGML_OPENBLAS=1
+    ```
+
+  - On Windows:
+
+    1. Download the latest fortran version of [w64devkit](https://github.com/skeeto/w64devkit/releases).
+    2. Download the latest version of [OpenBLAS for Windows](https://github.com/xianyi/OpenBLAS/releases).
+    3. Extract `w64devkit` on your pc.
+    4. From the OpenBLAS zip that you just downloaded copy `libopenblas.a`, located inside the `lib` folder, inside `w64devkit\x86_64-w64-mingw32\lib`.
+    5. From the same OpenBLAS zip copy the content of the `include` folder inside `w64devkit\x86_64-w64-mingw32\include`.
+    6. Run `w64devkit.exe`.
+    7. Use the `cd` command to reach the `llama.cpp` folder.
+    8. From here you can run:
+
+        ```bash
+        make GGML_OPENBLAS=1
+        ```
+
+- Using `CMake` on Linux:
+
+    ```bash
+    cmake -B build -DGGML_BLAS=ON -DGGML_BLAS_VENDOR=OpenBLAS
+    cmake --build build --config Release
+    ```
+
+### BLIS
+
+Check [BLIS.md](./backend/BLIS.md) for more information.
+
+### SYCL
+
+SYCL is a higher-level programming model to improve programming productivity on various hardware accelerators.
+
+llama.cpp based on SYCL is used to **support Intel GPU** (Data Center Max series, Flex series, Arc series, Built-in GPU and iGPU).
+
+For detailed info, please refer to [llama.cpp for SYCL](./backend/SYCL.md).
+
+### Intel oneMKL
+
+Building through oneAPI compilers will make avx_vnni instruction set available for intel processors that do not support avx512 and avx512_vnni. Please note that this build config **does not support Intel GPU**. For Intel GPU support, please refer to [llama.cpp for SYCL](./backend/SYCL.md).
+
+- Using manual oneAPI installation:
+  By default, `GGML_BLAS_VENDOR` is set to `Generic`, so if you already sourced intel environment script and assign `-DGGML_BLAS=ON` in cmake, the mkl version of Blas will automatically been selected. Otherwise please install oneAPI and follow the below steps:
+    ```bash
+    source /opt/intel/oneapi/setvars.sh # You can skip this step if  in oneapi-basekit docker image, only required for manual installation
+    cmake -B build -DGGML_BLAS=ON -DGGML_BLAS_VENDOR=Intel10_64lp -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx -DGGML_NATIVE=ON
+    cmake --build build --config Release
+    ```
+
+- Using oneAPI docker image:
+  If you do not want to source the environment vars and install oneAPI manually, you can also build the code using intel docker container: [oneAPI-basekit](https://hub.docker.com/r/intel/oneapi-basekit). Then, you can use the commands given above.
+
+Check [Optimizing and Running LLaMA2 on Intel® CPU](https://www.intel.com/content/www/us/en/content-details/791610/optimizing-and-running-llama2-on-intel-cpu.html) for more information.
+
+### CUDA
+
+This provides GPU acceleration using the CUDA cores of your Nvidia GPU. Make sure to have the CUDA toolkit installed. You can download it from your Linux distro's package manager (e.g. `apt install nvidia-cuda-toolkit`) or from here: [CUDA Toolkit](https://developer.nvidia.com/cuda-downloads).
+
+For Jetson user, if you have Jetson Orin, you can try this: [Offical Support](https://www.jetson-ai-lab.com/tutorial_text-generation.html). If you are using an old model(nano/TX2), need some additional operations before compiling.
+
+- Using `make`:
+  ```bash
+  make GGML_CUDA=1
+  ```
+- Using `CMake`:
+
+  ```bash
+  cmake -B build -DGGML_CUDA=ON
+  cmake --build build --config Release
+  ```
+
+The environment variable [`CUDA_VISIBLE_DEVICES`](https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#env-vars) can be used to specify which GPU(s) will be used. The following compilation options are also available to tweak performance:
+
+| Option                        | Legal values           | Default | Description                                                                                                                                                                                                                                                                             |
+|-------------------------------|------------------------|---------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
+| GGML_CUDA_FORCE_DMMV          | Boolean                | false   | Force the use of dequantization + matrix vector multiplication kernels instead of using kernels that do matrix vector multiplication on quantized data. By default the decision is made based on compute capability (MMVQ for 6.1/Pascal/GTX 1000 or higher). Does not affect k-quants. |
+| GGML_CUDA_DMMV_X              | Positive integer >= 32 | 32      | Number of values in x direction processed by the CUDA dequantization + matrix vector multiplication kernel per iteration. Increasing this value can improve performance on fast GPUs. Power of 2 heavily recommended. Does not affect k-quants.                                         |
+| GGML_CUDA_MMV_Y               | Positive integer       | 1       | Block size in y direction for the CUDA mul mat vec kernels. Increasing this value can improve performance on fast GPUs. Power of 2 recommended.                                                                                                                                         |
+| GGML_CUDA_FORCE_MMQ           | Boolean                | false   | Force the use of custom matrix multiplication kernels for quantized models instead of FP16 cuBLAS even if there is no int8 tensor core implementation available (affects V100, RDNA3). MMQ kernels are enabled by default on GPUs with int8 tensor core support. With MMQ force enabled, speed for large batch sizes will be worse but VRAM consumption will be lower.                       |
+| GGML_CUDA_FORCE_CUBLAS        | Boolean                | false   | Force the use of FP16 cuBLAS instead of custom matrix multiplication kernels for quantized models                                                                                                                                                                                       |
+| GGML_CUDA_F16                 | Boolean                | false   | If enabled, use half-precision floating point arithmetic for the CUDA dequantization + mul mat vec kernels and for the q4_1 and q5_1 matrix matrix multiplication kernels. Can improve performance on relatively recent GPUs.                                                           |
+| GGML_CUDA_KQUANTS_ITER        | 1 or 2                 | 2       | Number of values processed per iteration and per CUDA thread for Q2_K and Q6_K quantization formats. Setting this value to 1 can improve performance for slow GPUs.                                                                                                                     |
+| GGML_CUDA_PEER_MAX_BATCH_SIZE | Positive integer       | 128     | Maximum batch size for which to enable peer access between multiple GPUs. Peer access requires either Linux or NVLink. When using NVLink enabling peer access for larger batch sizes is potentially beneficial.                                                                         |
+| GGML_CUDA_FA_ALL_QUANTS       | Boolean                | false   | Compile support for all KV cache quantization type (combinations) for the FlashAttention CUDA kernels. More fine-grained control over KV cache size but compilation takes much longer.                                                                                                  |
+
+### hipBLAS
+
+This provides BLAS acceleration on HIP-supported AMD GPUs.
+Make sure to have ROCm installed.
+You can download it from your Linux distro's package manager or from here: [ROCm Quick Start (Linux)](https://rocm.docs.amd.com/projects/install-on-linux/en/latest/tutorial/quick-start.html#rocm-install-quick).
+
+- Using `make`:
+  ```bash
+  make GGML_HIPBLAS=1
+  ```
+- Using `CMake` for Linux (assuming a gfx1030-compatible AMD GPU):
+  ```bash
+  HIPCXX="$(hipconfig -l)/clang" HIP_PATH="$(hipconfig -R)" \
+      cmake -S . -B build -DGGML_HIPBLAS=ON -DAMDGPU_TARGETS=gfx1030 -DCMAKE_BUILD_TYPE=Release \
+      && cmake --build build --config Release -- -j 16
+  ```
+  On Linux it is also possible to use unified memory architecture (UMA) to share main memory between the CPU and integrated GPU by setting `-DGGML_HIP_UMA=ON`.
+  However, this hurts performance for non-integrated GPUs (but enables working with integrated GPUs).
+
+  Note that if you get the following error:
+  ```
+  clang: error: cannot find ROCm device library; provide its path via '--rocm-path' or '--rocm-device-lib-path', or pass '-nogpulib' to build without ROCm device library
+  ```
+  Try searching for a directory under `HIP_PATH` that contains the file
+  `oclc_abi_version_400.bc`. Then, add the following to the start of the
+  command: `HIP_DEVICE_LIB_PATH=<directory-you-just-found>`, so something
+  like:
+  ```bash
+  HIPCXX="$(hipconfig -l)/clang" HIP_PATH="$(hipconfig -p)" \
+  HIP_DEVICE_LIB_PATH=<directory-you-just-found> \
+      cmake -S . -B build -DGGML_HIPBLAS=ON -DAMDGPU_TARGETS=gfx1030 -DCMAKE_BUILD_TYPE=Release \
+      && cmake --build build -- -j 16
+  ```
+
+- Using `make` (example for target gfx1030, build with 16 CPU threads):
+  ```bash
+  make -j16 GGML_HIPBLAS=1 GGML_HIP_UMA=1 AMDGPU_TARGETS=gfx1030
+  ```
+
+- Using `CMake` for Windows (using x64 Native Tools Command Prompt for VS, and assuming a gfx1100-compatible AMD GPU):
+  ```bash
+  set PATH=%HIP_PATH%\bin;%PATH%
+  cmake -S . -B build -G Ninja -DAMDGPU_TARGETS=gfx1100 -DGGML_HIPBLAS=ON -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ -DCMAKE_BUILD_TYPE=Release
+  cmake --build build
+  ```
+  Make sure that `AMDGPU_TARGETS` is set to the GPU arch you want to compile for. The above example uses `gfx1100` that corresponds to Radeon RX 7900XTX/XT/GRE. You can find a list of targets [here](https://llvm.org/docs/AMDGPUUsage.html#processors)
+  Find your gpu version string by matching the most significant version information from `rocminfo | grep gfx | head -1 | awk '{print $2}'` with the list of processors, e.g. `gfx1035` maps to `gfx1030`.
+
+
+The environment variable [`HIP_VISIBLE_DEVICES`](https://rocm.docs.amd.com/en/latest/understand/gpu_isolation.html#hip-visible-devices) can be used to specify which GPU(s) will be used.
+If your GPU is not officially supported you can use the environment variable [`HSA_OVERRIDE_GFX_VERSION`] set to a similar GPU, for example 10.3.0 on RDNA2 (e.g. gfx1030, gfx1031, or gfx1035) or 11.0.0 on RDNA3.
+The following compilation options are also available to tweak performance (yes, they refer to CUDA, not HIP, because it uses the same code as the cuBLAS version above):
+
+| Option                 | Legal values           | Default | Description                                                                                                                                                                                                                                    |
+|------------------------|------------------------|---------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
+| GGML_CUDA_DMMV_X       | Positive integer >= 32 | 32      | Number of values in x direction processed by the HIP dequantization + matrix vector multiplication kernel per iteration. Increasing this value can improve performance on fast GPUs. Power of 2 heavily recommended. Does not affect k-quants. |
+| GGML_CUDA_MMV_Y        | Positive integer       | 1       | Block size in y direction for the HIP mul mat vec kernels. Increasing this value can improve performance on fast GPUs. Power of 2 recommended. Does not affect k-quants.                                                                       |
+| GGML_CUDA_KQUANTS_ITER | 1 or 2                 | 2       | Number of values processed per iteration and per HIP thread for Q2_K and Q6_K quantization formats. Setting this value to 1 can improve performance for slow GPUs.                                                                             |
+
+### Vulkan
+
+**With docker**:
+
+You don't need to install Vulkan SDK. It will be installed inside the container.
+
+```sh
+# Build the image
+docker build -t llama-cpp-vulkan -f .devops/llama-cli-vulkan.Dockerfile .
+
+# Then, use it:
+docker run -it --rm -v "$(pwd):/app:Z" --device /dev/dri/renderD128:/dev/dri/renderD128 --device /dev/dri/card1:/dev/dri/card1 llama-cpp-vulkan -m "/app/models/YOUR_MODEL_FILE" -p "Building a website can be done in 10 simple steps:" -n 400 -e -ngl 33
+```
+
+**Without docker**:
+
+Firstly, you need to make sure you have installed [Vulkan SDK](https://vulkan.lunarg.com/doc/view/latest/linux/getting_started_ubuntu.html)
+
+For example, on Ubuntu 22.04 (jammy), use the command below:
+
+```bash
+wget -qO - https://packages.lunarg.com/lunarg-signing-key-pub.asc | apt-key add -
+wget -qO /etc/apt/sources.list.d/lunarg-vulkan-jammy.list https://packages.lunarg.com/vulkan/lunarg-vulkan-jammy.list
+apt update -y
+apt-get install -y vulkan-sdk
+# To verify the installation, use the command below:
+vulkaninfo
+```
+
+Alternatively your package manager might be able to provide the appropriate libraries.
+For example for Ubuntu 22.04 you can install `libvulkan-dev` instead.
+For Fedora 40, you can install `vulkan-devel`, `glslc` and `glslang` packages.
+
+Then, build llama.cpp using the cmake command below:
+
+```bash
+cmake -B build -DGGML_VULKAN=1
+cmake --build build --config Release
+# Test the output binary (with "-ngl 33" to offload all layers to GPU)
+./bin/llama-cli -m "PATH_TO_MODEL" -p "Hi you how are you" -n 50 -e -ngl 33 -t 4
+
+# You should see in the output, ggml_vulkan detected your GPU. For example:
+# ggml_vulkan: Using Intel(R) Graphics (ADL GT2) | uma: 1 | fp16: 1 | warp size: 32
+```
+
+### Android
+
+To read documentation for how to build on Android, [click here](./android.md)
@@ -1,4 +1,4 @@
-## Add a new model architecture to `llama.cpp`
+# Add a new model architecture to `llama.cpp`

 Adding a model requires few steps:

@@ -17,7 +17,7 @@ Also, it is important to check that the examples and main ggml backends (CUDA, M
 ### 1. Convert the model to GGUF

 This step is done in python with a `convert` script using the [gguf](https://pypi.org/project/gguf/) library.
-Depending on the model architecture, you can use either [convert-hf-to-gguf.py](../convert-hf-to-gguf.py) or [examples/convert-legacy-llama.py](../examples/convert-legacy-llama.py) (for `llama/llama2` models in `.pth` format).
+Depending on the model architecture, you can use either [convert_hf_to_gguf.py](../convert_hf_to_gguf.py) or [examples/convert_legacy_llama.py](../examples/convert_legacy_llama.py) (for `llama/llama2` models in `.pth` format).

 The convert script reads the model configuration, tokenizer, tensor names+data and converts them to GGUF metadata and tensors.

@@ -0,0 +1,86 @@
+# Docker
+
+## Prerequisites
+* Docker must be installed and running on your system.
+* Create a folder to store big models & intermediate files (ex. /llama/models)
+
+## Images
+We have three Docker images available for this project:
+
+1. `ghcr.io/ggerganov/llama.cpp:full`: This image includes both the main executable file and the tools to convert LLaMA models into ggml and convert into 4-bit quantization. (platforms: `linux/amd64`, `linux/arm64`)
+2. `ghcr.io/ggerganov/llama.cpp:light`: This image only includes the main executable file. (platforms: `linux/amd64`, `linux/arm64`)
+3. `ghcr.io/ggerganov/llama.cpp:server`: This image only includes the server executable file. (platforms: `linux/amd64`, `linux/arm64`)
+
+Additionally, there the following images, similar to the above:
+
+- `ghcr.io/ggerganov/llama.cpp:full-cuda`: Same as `full` but compiled with CUDA support. (platforms: `linux/amd64`)
+- `ghcr.io/ggerganov/llama.cpp:light-cuda`: Same as `light` but compiled with CUDA support. (platforms: `linux/amd64`)
+- `ghcr.io/ggerganov/llama.cpp:server-cuda`: Same as `server` but compiled with CUDA support. (platforms: `linux/amd64`)
+- `ghcr.io/ggerganov/llama.cpp:full-rocm`: Same as `full` but compiled with ROCm support. (platforms: `linux/amd64`, `linux/arm64`)
+- `ghcr.io/ggerganov/llama.cpp:light-rocm`: Same as `light` but compiled with ROCm support. (platforms: `linux/amd64`, `linux/arm64`)
+- `ghcr.io/ggerganov/llama.cpp:server-rocm`: Same as `server` but compiled with ROCm support. (platforms: `linux/amd64`, `linux/arm64`)
+
+The GPU enabled images are not currently tested by CI beyond being built. They are not built with any variation from the ones in the Dockerfiles defined in [.devops/](.devops/) and the GitHub Action defined in [.github/workflows/docker.yml](.github/workflows/docker.yml). If you need different settings (for example, a different CUDA or ROCm library, you'll need to build the images locally for now).
+
+## Usage
+
+The easiest way to download the models, convert them to ggml and optimize them is with the --all-in-one command which includes the full docker image.
+
+Replace `/path/to/models` below with the actual path where you downloaded the models.
+
+```bash
+docker run -v /path/to/models:/models ghcr.io/ggerganov/llama.cpp:full --all-in-one "/models/" 7B
+```
+
+On completion, you are ready to play!
+
+```bash
+docker run -v /path/to/models:/models ghcr.io/ggerganov/llama.cpp:full --run -m /models/7B/ggml-model-q4_0.gguf -p "Building a website can be done in 10 simple steps:" -n 512
+```
+
+or with a light image:
+
+```bash
+docker run -v /path/to/models:/models ghcr.io/ggerganov/llama.cpp:light -m /models/7B/ggml-model-q4_0.gguf -p "Building a website can be done in 10 simple steps:" -n 512
+```
+
+or with a server image:
+
+```bash
+docker run -v /path/to/models:/models -p 8000:8000 ghcr.io/ggerganov/llama.cpp:server -m /models/7B/ggml-model-q4_0.gguf --port 8000 --host 0.0.0.0 -n 512
+```
+
+## Docker With CUDA
+
+Assuming one has the [nvidia-container-toolkit](https://github.com/NVIDIA/nvidia-container-toolkit) properly installed on Linux, or is using a GPU enabled cloud, `cuBLAS` should be accessible inside the container.
+
+## Building Docker locally
+
+```bash
+docker build -t local/llama.cpp:full-cuda -f .devops/full-cuda.Dockerfile .
+docker build -t local/llama.cpp:light-cuda -f .devops/llama-cli-cuda.Dockerfile .
+docker build -t local/llama.cpp:server-cuda -f .devops/llama-server-cuda.Dockerfile .
+```
+
+You may want to pass in some different `ARGS`, depending on the CUDA environment supported by your container host, as well as the GPU architecture.
+
+The defaults are:
+
+- `CUDA_VERSION` set to `11.7.1`
+- `CUDA_DOCKER_ARCH` set to `all`
+
+The resulting images, are essentially the same as the non-CUDA images:
+
+1. `local/llama.cpp:full-cuda`: This image includes both the main executable file and the tools to convert LLaMA models into ggml and convert into 4-bit quantization.
+2. `local/llama.cpp:light-cuda`: This image only includes the main executable file.
+3. `local/llama.cpp:server-cuda`: This image only includes the server executable file.
+
+## Usage
+
+After building locally, Usage is similar to the non-CUDA examples, but you'll need to add the `--gpus` flag. You will also want to use the `--n-gpu-layers` flag.
+
+```bash
+docker run --gpus all -v /path/to/models:/models local/llama.cpp:full-cuda --run -m /models/7B/ggml-model-q4_0.gguf -p "Building a website can be done in 10 simple steps:" -n 512 --n-gpu-layers 1
+docker run --gpus all -v /path/to/models:/models local/llama.cpp:light-cuda -m /models/7B/ggml-model-q4_0.gguf -p "Building a website can be done in 10 simple steps:" -n 512 --n-gpu-layers 1
+docker run --gpus all -v /path/to/models:/models local/llama.cpp:server-cuda -m /models/7B/ggml-model-q4_0.gguf --port 8000 --host 0.0.0.0 -n 512 --n-gpu-layers 1
+```
@@ -0,0 +1,39 @@
+# Install pre-built version of llama.cpp
+
+## Homebrew
+
+On Mac and Linux, the homebrew package manager can be used via
+
+```sh
+brew install llama.cpp
+```
+The formula is automatically updated with new `llama.cpp` releases. More info: https://github.com/ggerganov/llama.cpp/discussions/7668
+
+## Nix
+
+On Mac and Linux, the Nix package manager can be used via
+
+```sh
+nix profile install nixpkgs#llama-cpp
+```
+For flake enabled installs.
+
+Or
+
+```sh
+nix-env --file '<nixpkgs>' --install --attr llama-cpp
+```
+
+For non-flake enabled installs.
+
+This expression is automatically updated within the [nixpkgs repo](https://github.com/NixOS/nixpkgs/blob/nixos-24.05/pkgs/by-name/ll/llama-cpp/package.nix#L164).
+
+## Flox
+
+On Mac and Linux, Flox can be used to install llama.cpp within a Flox environment via
+
+```sh
+flox install llama-cpp
+```
+
+Flox follows the nixpkgs build of llama.cpp.
@@ -229,7 +229,7 @@ private func tokenize(text: String, add_bos: Bool) -> [llama_token] {

 private func token_to_piece(token: llama_token, buffer: inout [CChar]) -> String? {
    var result = [CChar](repeating: 0, count: 8)
-    let nTokens = llama_token_to_piece(model, token, &result, Int32(result.count), false)
+    let nTokens = llama_token_to_piece(model, token, &result, Int32(result.count), 0, false)
    if nTokens < 0 {
        let actualTokensCount = -Int(nTokens)
        result = .init(repeating: 0, count: actualTokensCount)
@@ -238,6 +238,7 @@ private func token_to_piece(token: llama_token, buffer: inout [CChar]) -> String
            token,
            &result,
            Int32(result.count),
+            0,
            false
        )
        assert(check == actualTokensCount)
@@ -93,14 +93,34 @@ int main(int argc, char ** argv) {

    // create a llama_batch
    // we use this object to submit token data for decoding
-    llama_batch batch = llama_batch_init(std::max(tokens_list.size(), (size_t)n_parallel), 0, 1);
+    llama_batch batch = llama_batch_init(std::max(tokens_list.size(), (size_t) n_parallel), 0, n_parallel);
+
+    std::vector<llama_seq_id> seq_ids(n_parallel, 0);
+    for (int32_t i = 0; i < n_parallel; ++i) {
+        seq_ids[i] = i;
+    }

    // evaluate the initial prompt
    for (size_t i = 0; i < tokens_list.size(); ++i) {
-        llama_batch_add(batch, tokens_list[i], i, { 0 }, false);
+        llama_batch_add(batch, tokens_list[i], i, seq_ids, false);
    }
    GGML_ASSERT(batch.n_tokens == (int) tokens_list.size());

+    if (llama_model_has_encoder(model)) {
+        if (llama_encode(ctx, batch)) {
+            LOG_TEE("%s : failed to eval\n", __func__);
+            return 1;
+        }
+
+        llama_token decoder_start_token_id = llama_model_decoder_start_token(model);
+        if (decoder_start_token_id == -1) {
+            decoder_start_token_id = llama_token_bos(model);
+        }
+
+        llama_batch_clear(batch);
+        llama_batch_add(batch, decoder_start_token_id, 0, seq_ids, false);
+    }
+
    // llama_decode will output logits only for the last token of the prompt
    batch.logits[batch.n_tokens - 1] = true;

@@ -109,11 +129,11 @@ int main(int argc, char ** argv) {
        return 1;
    }

-    // assign the system KV cache to all parallel sequences
-    // this way, the parallel sequences will "reuse" the prompt tokens without having to copy them
-    for (int32_t i = 1; i < n_parallel; ++i) {
-        llama_kv_cache_seq_cp(ctx, 0, i, -1, -1);
-    }
+    //// assign the system KV cache to all parallel sequences
+    //// this way, the parallel sequences will "reuse" the prompt tokens without having to copy them
+    //for (int32_t i = 1; i < n_parallel; ++i) {
+    //    llama_kv_cache_seq_cp(ctx, 0, i, -1, -1);
+    //}

    if (n_parallel > 1) {
        LOG_TEE("\n\n%s: generating %d sequences ...\n", __func__, n_parallel);
@@ -58,4 +58,3 @@ The above command will output space-separated float values.
 ```powershell
 embedding.exe -p 'Castle<#sep#>Stronghold<#sep#>Dog<#sep#>Cat' --embd-separator '<#sep#>' --embd-normalize 2  --embd-output-format '' -m './path/to/model.gguf' --n-gpu-layers 99 --log-disable 2>/dev/null
 ```
-
@@ -87,4 +87,4 @@ The LORA rank can be configured for each model tensor type separately with these

 The LORA rank of 'norm' tensors should always be 1.

-To see all available options use `finetune --help`.
+To see all available options use `llama-finetune --help`.
@@ -8,7 +8,7 @@ if [[ ! $LLAMA_MODEL_DIR ]]; then LLAMA_MODEL_DIR="./models"; fi
 if [[ ! $LLAMA_TRAINING_DIR ]]; then LLAMA_TRAINING_DIR="."; fi

 # MODEL="$LLAMA_MODEL_DIR/openllama-3b-v2-q8_0.gguf" # This is the model the readme uses.
-MODEL="$LLAMA_MODEL_DIR/openllama-3b-v2.gguf" # An f16 model. Note in this case with "-g", you get an f32-format .BIN file that isn't yet supported if you use it with "main --lora" with GPU inferencing.
+MODEL="$LLAMA_MODEL_DIR/openllama-3b-v2.gguf" # An f16 model. Note in this case with "-g", you get an f32-format .BIN file that isn't yet supported if you use it with "llama-cli --lora" with GPU inferencing.

 while getopts "dg" opt; do
  case $opt in
@@ -15,6 +15,7 @@ In this section, we cover the most commonly used options for running the `infill
 -   `-i, --interactive`: Run the program in interactive mode, allowing you to provide input directly and receive real-time responses.
 -   `-n N, --n-predict N`: Set the number of tokens to predict when generating text. Adjusting this value can influence the length of the generated text.
 -   `-c N, --ctx-size N`: Set the size of the prompt context. The default is 512, but LLaMA models were built with a context of 2048, which will provide better results for longer input/inference.
+-   `--spm-infill`: Use Suffix/Prefix/Middle pattern for infill (instead of Prefix/Suffix/Middle) as some models prefer this.

 ## Input Prompts

@@ -210,6 +210,7 @@ int main(int argc, char ** argv) {
        suff_rm_leading_spc = false;
    }
    std::vector<llama_token> embd_inp;
+    std::vector<llama_token> embd_end;
    std::vector<llama_token> inp_pfx = ::llama_tokenize(ctx, params.input_prefix, false);
    std::vector<llama_token> inp_sfx = ::llama_tokenize(ctx, params.input_suffix, false);
    const int space_token = 29871;
@@ -217,12 +218,13 @@ int main(int argc, char ** argv) {
        inp_sfx.erase(inp_sfx.begin());
    }
    inp_pfx.insert(inp_pfx.begin(), llama_token_prefix(model));
-    if (add_bos) {
-        inp_pfx.insert(inp_pfx.begin(), llama_token_bos(model));
-    }
    inp_sfx.insert(inp_sfx.begin(), llama_token_suffix(model));
-    embd_inp = inp_pfx;
-    embd_inp.insert(embd_inp.end(), inp_sfx.begin(), inp_sfx.end());
+    embd_inp = params.spm_infill ? inp_sfx : inp_pfx;
+    embd_end = params.spm_infill ? inp_pfx : inp_sfx;
+    if (add_bos) {
+        embd_inp.insert(embd_inp.begin(), llama_token_bos(model));
+    }
+    embd_inp.insert(embd_inp.end(), embd_end.begin(), embd_end.end());

    const llama_token middle_token = llama_token_middle(model);
    if (middle_token >= 0) {
@@ -526,14 +528,14 @@ int main(int argc, char ** argv) {
                    inp_sfx.erase(inp_sfx.begin());
                }
                inp_pfx.insert(inp_pfx.begin(), llama_token_prefix(model));
-                if (add_bos) {
-                    inp_pfx.insert(inp_pfx.begin(), llama_token_bos(model));
-                }
                inp_sfx.insert(inp_sfx.begin(), llama_token_suffix(model));
-                embd_inp = inp_pfx;
-                embd_inp.insert(embd_inp.end(), inp_sfx.begin(), inp_sfx.end());
+                embd_inp = params.spm_infill ? inp_sfx : inp_pfx;
+                embd_end = params.spm_infill ? inp_pfx : inp_sfx;
+                if (add_bos) {
+                    embd_inp.insert(embd_inp.begin(), llama_token_bos(model));
+                }
+                embd_inp.insert(embd_inp.end(), embd_end.begin(), embd_end.end());

-                const llama_token middle_token = llama_token_middle(model);
                if (middle_token >= 0) {
                    embd_inp.push_back(middle_token);
                }
@@ -657,4 +659,3 @@ int main(int argc, char ** argv) {

    return 0;
 }
-
@@ -1,7 +1,7 @@
 # Usage:
 #! ./llama-server -m some-model.gguf &
 #! pip install pydantic
-#! python json-schema-pydantic-example.py
+#! python json_schema_pydantic_example.py

 from pydantic import BaseModel, Extra, TypeAdapter
 from annotated_types import MinLen
@@ -231,7 +231,7 @@ GRAMMAR_RANGE_LITERAL_ESCAPE_RE = re.compile(r'[\r\n"\]\-\\]')
 GRAMMAR_LITERAL_ESCAPES = {'\r': '\\r', '\n': '\\n', '"': '\\"', '-': '\\-', ']': '\\]'}

 NON_LITERAL_SET = set('|.()[]{}*+?')
-ESCAPED_IN_REGEXPS_BUT_NOT_IN_LITERALS = set('[]()|{}*+?')
+ESCAPED_IN_REGEXPS_BUT_NOT_IN_LITERALS = set('^$.[]()|{}*+?')


 class SchemaConverter:
@@ -602,7 +602,7 @@ class SchemaConverter:
                else:
                    add_component(t, is_required=True)

-            return self._add_rule(rule_name, self._build_object_rule(properties, required, hybrid_name, additional_properties=[]))
+            return self._add_rule(rule_name, self._build_object_rule(properties, required, hybrid_name, additional_properties=None))

        elif schema_type in (None, 'array') and ('items' in schema or 'prefixItems' in schema):
            items = schema.get('items') or schema['prefixItems']
@@ -691,7 +691,7 @@ class SchemaConverter:
        required_props = [k for k in sorted_props if k in required]
        optional_props = [k for k in sorted_props if k not in required]

-        if additional_properties != False:
+        if additional_properties is not None and additional_properties != False:
            sub_name = f'{name}{"-" if name else ""}additional'
            value_rule = self.visit(additional_properties, f'{sub_name}-value') if isinstance(additional_properties, dict) else \
                self._add_primitive('value', PRIMITIVE_RULES['value'])
@@ -1,55 +0,0 @@
-
-# For more information about using CMake with Android Studio, read the
-# documentation: https://d.android.com/studio/projects/add-native-code.html.
-# For more examples on how to use CMake, see https://github.com/android/ndk-samples.
-
-# Sets the minimum CMake version required for this project.
-cmake_minimum_required(VERSION 3.22.1)
-
-# Declares the project name. The project name can be accessed via ${ PROJECT_NAME},
-# Since this is the top level CMakeLists.txt, the project name is also accessible
-# with ${CMAKE_PROJECT_NAME} (both CMake variables are in-sync within the top level
-# build script scope).
-project("llama-android")
-
-## Fetch latest llama.cpp from GitHub
-#include(FetchContent)
-#FetchContent_Declare(
-#        llama
-#        GIT_REPOSITORY https://github.com/ggerganov/llama.cpp
-#        GIT_TAG        master
-#)
-#
-## Also provides "common"
-#FetchContent_MakeAvailable(llama)
-
-# llama.cpp CI uses the code from the current branch
-# ref: https://github.com/ggerganov/llama.cpp/pull/7341#issuecomment-2117617700
-add_subdirectory(../../../../../../ build-llama)
-
-# Creates and names a library, sets it as either STATIC
-# or SHARED, and provides the relative paths to its source code.
-# You can define multiple libraries, and CMake builds them for you.
-# Gradle automatically packages shared libraries with your APK.
-#
-# In this top level CMakeLists.txt, ${CMAKE_PROJECT_NAME} is used to define
-# the target library name; in the sub-module's CMakeLists.txt, ${PROJECT_NAME}
-# is preferred for the same purpose.
-#
-# In order to load a library into your app from Java/Kotlin, you must call
-# System.loadLibrary() and pass the name of the library defined here;
-# for GameActivity/NativeActivity derived applications, the same library name must be
-# used in the AndroidManifest.xml file.
-add_library(${CMAKE_PROJECT_NAME} SHARED
-    # List C/C++ source files with relative paths to this CMakeLists.txt.
-        llama-android.cpp)
-
-# Specifies libraries CMake should link to your target library. You
-# can link libraries from various origins, such as libraries defined in this
-# build script, prebuilt third-party libraries, or Android system libraries.
-target_link_libraries(${CMAKE_PROJECT_NAME}
-    # List libraries link to the target library
-    llama
-    common
-    android
-    log)
@@ -11,15 +11,15 @@ cmake_minimum_required(VERSION 3.22.1)
 # build script scope).
 project("llama-android")

-include(FetchContent)
-FetchContent_Declare(
-        llama
-        GIT_REPOSITORY https://github.com/ggerganov/llama.cpp
-        GIT_TAG        master
-)
+#include(FetchContent)
+#FetchContent_Declare(
+#        llama
+#        GIT_REPOSITORY https://github.com/ggerganov/llama.cpp
+#        GIT_TAG        master
+#)

 # Also provides "common"
-FetchContent_MakeAvailable(llama)
+#FetchContent_MakeAvailable(llama)

 # Creates and names a library, sets it as either STATIC
 # or SHARED, and provides the relative paths to its source code.
@@ -30,6 +30,10 @@ FetchContent_MakeAvailable(llama)
 # the target library name; in the sub-module's CMakeLists.txt, ${PROJECT_NAME}
 # is preferred for the same purpose.
 #
+
+#load local llama.cpp
+add_subdirectory(../../../../../../ build-llama)
+
 # In order to load a library into your app from Java/Kotlin, you must call
 # System.loadLibrary() and pass the name of the library defined here;
 # for GameActivity/NativeActivity derived applications, the same library name must be
@@ -5,7 +5,7 @@
 #include <string>
 #include <unistd.h>
 #include "llama.h"
-#include "common/common.h"
+#include "common.h"

 // Write C++ code here.
 //
@@ -322,7 +322,7 @@ actor LlamaContext {
        defer {
            result.deallocate()
        }
-        let nTokens = llama_token_to_piece(model, token, result, 8, false)
+        let nTokens = llama_token_to_piece(model, token, result, 8, 0, false)

        if nTokens < 0 {
            let newResult = UnsafeMutablePointer<Int8>.allocate(capacity: Int(-nTokens))
@@ -330,7 +330,7 @@ actor LlamaContext {
            defer {
                newResult.deallocate()
            }
-            let nNewTokens = llama_token_to_piece(model, token, newResult, -nTokens, false)
+            let nNewTokens = llama_token_to_piece(model, token, newResult, -nTokens, 0, false)
            let bufferPointer = UnsafeBufferPointer(start: newResult, count: Int(nNewTokens))
            return Array(bufferPointer)
        } else {
@@ -30,16 +30,16 @@ git clone https://huggingface.co/mtgv/MobileVLM-1.7B
 git clone https://huggingface.co/openai/clip-vit-large-patch14-336
 ```

-2. Use `llava-surgery.py` to split the LLaVA model to LLaMA and multimodel projector constituents:
+2. Use `llava_surgery.py` to split the LLaVA model to LLaMA and multimodel projector constituents:

 ```sh
-python ./examples/llava/llava-surgery.py -m path/to/MobileVLM-1.7B
+python ./examples/llava/llava_surgery.py -m path/to/MobileVLM-1.7B
 ```

-3. Use `convert-image-encoder-to-gguf.py` with `--projector-type ldp` (for **V2** please use `--projector-type ldpv2`) to convert the LLaVA image encoder to GGUF:
+3. Use `convert_image_encoder_to_gguf.py` with `--projector-type ldp` (for **V2** please use `--projector-type ldpv2`) to convert the LLaVA image encoder to GGUF:

 ```sh
-python ./examples/llava/convert-image-encoder-to-gguf \
+python ./examples/llava/convert_image_encoder_to_gguf \
    -m path/to/clip-vit-large-patch14-336 \
    --llava-projector path/to/MobileVLM-1.7B/llava.projector \
    --output-dir path/to/MobileVLM-1.7B \
@@ -47,17 +47,17 @@ python ./examples/llava/convert-image-encoder-to-gguf \
 ```

 ```sh
-python ./examples/llava/convert-image-encoder-to-gguf \
+python ./examples/llava/convert_image_encoder_to_gguf \
    -m path/to/clip-vit-large-patch14-336 \
    --llava-projector path/to/MobileVLM-1.7B_V2/llava.projector \
    --output-dir path/to/MobileVLM-1.7B_V2 \
    --projector-type ldpv2
 ```

-4. Use `examples/convert-legacy-llama.py` to convert the LLaMA part of LLaVA to GGUF:
+4. Use `examples/convert_legacy_llama.py` to convert the LLaMA part of LLaVA to GGUF:

 ```sh
-python ./examples/convert-legacy-llama.py path/to/MobileVLM-1.7B
+python ./examples/convert_legacy_llama.py path/to/MobileVLM-1.7B
 ```

 5. Use `quantize` to convert LLaMA part's DataType from `fp16` to `q4_k`
@@ -38,22 +38,22 @@ git clone https://huggingface.co/openai/clip-vit-large-patch14-336
 pip install -r examples/llava/requirements.txt
 ```

-3. Use `llava-surgery.py` to split the LLaVA model to LLaMA and multimodel projector constituents:
+3. Use `llava_surgery.py` to split the LLaVA model to LLaMA and multimodel projector constituents:

 ```sh
-python ./examples/llava/llava-surgery.py -m ../llava-v1.5-7b
+python ./examples/llava/llava_surgery.py -m ../llava-v1.5-7b
 ```

-4. Use `convert-image-encoder-to-gguf.py` to convert the LLaVA image encoder to GGUF:
+4. Use `convert_image_encoder_to_gguf.py` to convert the LLaVA image encoder to GGUF:

 ```sh
-python ./examples/llava/convert-image-encoder-to-gguf.py -m ../clip-vit-large-patch14-336 --llava-projector ../llava-v1.5-7b/llava.projector --output-dir ../llava-v1.5-7b
+python ./examples/llava/convert_image_encoder_to_gguf.py -m ../clip-vit-large-patch14-336 --llava-projector ../llava-v1.5-7b/llava.projector --output-dir ../llava-v1.5-7b
 ```

-5. Use `examples/convert-legacy-llama.py` to convert the LLaMA part of LLaVA to GGUF:
+5. Use `examples/convert_legacy_llama.py` to convert the LLaMA part of LLaVA to GGUF:

 ```sh
-python ./examples/convert-legacy-llama.py ../llava-v1.5-7b --skip-unknown
+python ./examples/convert_legacy_llama.py ../llava-v1.5-7b --skip-unknown
 ```

 Now both the LLaMA part and the image encoder are in the `llava-v1.5-7b` directory.
@@ -70,9 +70,9 @@ git clone https://huggingface.co/liuhaotian/llava-v1.6-vicuna-7b
 pip install -r examples/llava/requirements.txt
 ```

-3) Use `llava-surgery-v2.py` which also supports llava-1.5 variants pytorch as well as safetensor models:
+3) Use `llava_surgery_v2.py` which also supports llava-1.5 variants pytorch as well as safetensor models:
 ```console
-python examples/llava/llava-surgery-v2.py -C -m ../llava-v1.6-vicuna-7b/
+python examples/llava/llava_surgery_v2.py -C -m ../llava-v1.6-vicuna-7b/
 ```
 - you will find a llava.projector and a llava.clip file in your model directory

@@ -86,13 +86,13 @@ curl -s -q https://huggingface.co/cmp-nct/llava-1.6-gguf/raw/main/config_vit.jso

 5) Create the visual gguf model:
 ```console
-python ./examples/llava/convert-image-encoder-to-gguf.py -m vit --llava-projector vit/llava.projector --output-dir vit --clip-model-is-vision
+python ./examples/llava/convert_image_encoder_to_gguf.py -m vit --llava-projector vit/llava.projector --output-dir vit --clip-model-is-vision
 ```
 - This is similar to llava-1.5, the difference is that we tell the encoder that we are working with the pure vision model part of CLIP

 6) Then convert the model to gguf format:
 ```console
-python ./examples/convert-legacy-llama.py ../llava-v1.6-vicuna-7b/ --skip-unknown
+python ./examples/convert_legacy_llama.py ../llava-v1.6-vicuna-7b/ --skip-unknown
 ```

 7) And finally we can run the llava cli using the 1.6 model version:
@@ -1121,20 +1121,20 @@ struct clip_ctx * clip_model_load(const char * fname, const int verbosity = 1) {
            }
            if (n < 32)
                hparams.image_grid_pinpoints[n] = 0;
-        } catch (std::runtime_error & e) {
+        } catch (std::runtime_error & /*e*/) {
            hparams.image_grid_pinpoints[0]=0;
        }

        try {
            int idx = get_key_idx(ctx, KEY_MM_PATCH_MERGE_TYPE);
            strcpy(hparams.mm_patch_merge_type, gguf_get_val_str(ctx, idx));
-        } catch (std::runtime_error & e) {
+        } catch (std::runtime_error & /*e*/) {
            strcpy(hparams.mm_patch_merge_type, "flat");
        }

        try {
            hparams.image_crop_resolution = get_u32(ctx, KEY_IMAGE_CROP_RESOLUTION); // llava-1.6
-        } catch(const std::exception& e) {
+        } catch(const std::exception& /*e*/) {
            hparams.image_crop_resolution = hparams.image_size;
        }

@@ -1173,7 +1173,7 @@ struct clip_ctx * clip_model_load(const char * fname, const int verbosity = 1) {
        try {
            vision_model.class_embedding  = get_tensor(new_clip->ctx_data, TN_CLASS_EMBD);
            new_clip->has_class_embedding = true;
-        } catch (const std::exception& e) {
+        } catch (const std::exception& /*e*/) {
            new_clip->has_class_embedding = false;
        }

@@ -1181,7 +1181,7 @@ struct clip_ctx * clip_model_load(const char * fname, const int verbosity = 1) {
            vision_model.pre_ln_w  = get_tensor(new_clip->ctx_data, format(TN_LN_PRE, "v", "weight"));
            vision_model.pre_ln_b  = get_tensor(new_clip->ctx_data, format(TN_LN_PRE, "v", "bias"));
            new_clip->has_pre_norm = true;
-        } catch (std::exception & e) {
+        } catch (std::exception & /*e*/) {
            new_clip->has_pre_norm = false;
        }

@@ -1189,21 +1189,21 @@ struct clip_ctx * clip_model_load(const char * fname, const int verbosity = 1) {
            vision_model.post_ln_w  = get_tensor(new_clip->ctx_data, format(TN_LN_POST, "v", "weight"));
            vision_model.post_ln_b  = get_tensor(new_clip->ctx_data, format(TN_LN_POST, "v", "bias"));
            new_clip->has_post_norm = true;
-        } catch (std::exception & e) {
+        } catch (std::exception & /*e*/) {
            new_clip->has_post_norm = false;
        }

        try {
            vision_model.patch_bias = get_tensor(new_clip->ctx_data, TN_PATCH_BIAS);
            new_clip->has_patch_bias = true;
-        } catch (std::exception & e) {
+        } catch (std::exception & /*e*/) {
            new_clip->has_patch_bias = false;
        }

        try {
            vision_model.patch_embeddings    = get_tensor(new_clip->ctx_data, TN_PATCH_EMBD);
            vision_model.position_embeddings = get_tensor(new_clip->ctx_data, format(TN_POS_EMBD, "v"));
-        } catch(const std::exception& e) {
+        } catch(const std::exception& /*e*/) {
            LOG_TEE("%s: failed to load vision model tensors\n", __func__);
        }

@@ -1215,26 +1215,26 @@ struct clip_ctx * clip_model_load(const char * fname, const int verbosity = 1) {
                // Yi-type llava
                vision_model.mm_1_w = get_tensor(new_clip->ctx_data, format(TN_LLAVA_PROJ, 1, "weight"));
                vision_model.mm_1_b = get_tensor(new_clip->ctx_data, format(TN_LLAVA_PROJ, 1, "bias"));
-            } catch (std::runtime_error & e) {  }
+            } catch (std::runtime_error & /*e*/) { }
            try {
                // missing in Yi-type llava
                vision_model.mm_2_w              = get_tensor(new_clip->ctx_data, format(TN_LLAVA_PROJ, 2, "weight"));
                vision_model.mm_2_b              = get_tensor(new_clip->ctx_data, format(TN_LLAVA_PROJ, 2, "bias"));
-            } catch (std::runtime_error & e) {  }
+            } catch (std::runtime_error & /*e*/) { }
            try {
                // Yi-type llava
                vision_model.mm_3_w = get_tensor(new_clip->ctx_data, format(TN_LLAVA_PROJ, 3, "weight"));
                vision_model.mm_3_b = get_tensor(new_clip->ctx_data, format(TN_LLAVA_PROJ, 3, "bias"));
-            } catch (std::runtime_error & e) {  }
+            } catch (std::runtime_error & /*e*/) { }
            try {
                // Yi-type llava
                vision_model.mm_4_w = get_tensor(new_clip->ctx_data, format(TN_LLAVA_PROJ, 4, "weight"));
                vision_model.mm_4_b = get_tensor(new_clip->ctx_data, format(TN_LLAVA_PROJ, 4, "bias"));
-            } catch (std::runtime_error & e) {  }
+            } catch (std::runtime_error & /*e*/) { }
            try {
                vision_model.image_newline = get_tensor(new_clip->ctx_data, TN_IMAGE_NEWLINE);
                // LOG_TEE("%s: image_newline tensor (llava-1.6) found\n", __func__);
-            } catch (std::runtime_error & e) {  }
+            } catch (std::runtime_error & /*e*/) { }
        } else if (new_clip->proj_type == PROJECTOR_TYPE_LDP) {
            // MobileVLM projection
            vision_model.mm_model_mlp_1_w               = get_tensor(new_clip->ctx_data, format(TN_MVLM_PROJ_MLP, 1, "weight"));
@@ -1,3 +1,4 @@
-r ../../requirements/requirements-convert-legacy-llama.txt
+-r ../../requirements/requirements-convert_legacy_llama.txt
+--extra-index-url https://download.pytorch.org/whl/cpu
 pillow~=10.2.0
-torch~=2.1.1
+torch~=2.2.1
@@ -10,4 +10,3 @@ More info:

 https://github.com/ggerganov/llama.cpp/pull/4484
 https://github.com/ggerganov/llama.cpp/issues/4226
-
@@ -48,4 +48,3 @@
 build*/
 out/
 tmp/
-
@@ -30,4 +30,3 @@ target_include_directories(${TARGET} PRIVATE ${_common_path})
 install(TARGETS ${TARGET} RUNTIME)
 target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
 target_compile_features(${TARGET} PRIVATE cxx_std_11)
-
@@ -37,7 +37,8 @@ static gpt_params               * g_params;
 static std::vector<llama_token> * g_input_tokens;
 static std::ostringstream       * g_output_ss;
 static std::vector<llama_token> * g_output_tokens;
-static bool is_interacting = false;
+static bool is_interacting  = false;
+static bool need_insert_eot = false;

 static bool file_exists(const std::string & path) {
    std::ifstream f(path.c_str());
@@ -99,7 +100,8 @@ static void write_logfile(
 static void sigint_handler(int signo) {
    if (signo == SIGINT) {
        if (!is_interacting && g_params->interactive) {
-            is_interacting = true;
+            is_interacting  = true;
+            need_insert_eot = true;
        } else {
            console::cleanup();
            printf("\n");
@@ -224,7 +226,14 @@ int main(int argc, char ** argv) {
                __func__, n_ctx_train, n_ctx);
    }

-    LOG_TEE("%s: chat template example: %s\n", __func__, llama_chat_format_example(model, params.chat_template).c_str());
+    // print chat template example in conversation mode
+    if (params.conversation) {
+        if (params.enable_chat_template) {
+            LOG_TEE("%s: chat template example: %s\n", __func__, llama_chat_format_example(model, params.chat_template).c_str());
+        } else {
+            LOG_TEE("%s: in-suffix/prefix is specified, chat template will be disabled\n", __func__);
+        }
+    }

    // print system information
    {
@@ -255,13 +264,15 @@ int main(int argc, char ** argv) {
    }

    const bool add_bos = llama_should_add_bos_token(model);
-    GGML_ASSERT(llama_add_eos_token(model) != 1);
+    if (!llama_model_has_encoder(model)) {
+        GGML_ASSERT(llama_add_eos_token(model) != 1);
+    }
    LOG("add_bos: %d\n", add_bos);

    std::vector<llama_token> embd_inp;

    {
-        auto prompt = params.conversation
+        auto prompt = (params.conversation && params.enable_chat_template && !params.prompt.empty())
            ? chat_add_and_format(model, chat_msgs, "system", params.prompt) // format the system prompt in conversation mode
            : params.prompt;
        if (params.interactive_first || !params.prompt.empty() || session_tokens.empty()) {
@@ -517,6 +528,24 @@ int main(int argc, char ** argv) {
        exit(1);
    }

+    if (llama_model_has_encoder(model)) {
+        int enc_input_size = embd_inp.size();
+        llama_token * enc_input_buf = embd_inp.data();
+
+        if (llama_encode(ctx, llama_batch_get_one(enc_input_buf, enc_input_size, 0, 0))) {
+            LOG_TEE("%s : failed to eval\n", __func__);
+            return 1;
+        }
+
+        llama_token decoder_start_token_id = llama_model_decoder_start_token(model);
+        if (decoder_start_token_id == -1) {
+            decoder_start_token_id = llama_token_bos(model);
+        }
+
+        embd_inp.clear();
+        embd_inp.push_back(decoder_start_token_id);
+    }
+
    while ((n_remain != 0 && !is_antiprompt) || params.interactive) {
        // predict
        if (!embd.empty()) {
@@ -810,7 +839,9 @@ int main(int argc, char ** argv) {
                        is_antiprompt = true;
                    }

-                    chat_add_and_format(model, chat_msgs, "system", assistant_ss.str());
+                    if (params.enable_chat_template) {
+                        chat_add_and_format(model, chat_msgs, "assistant", assistant_ss.str());
+                    }
                    is_interacting = true;
                    printf("\n");
                }
@@ -872,16 +903,24 @@ int main(int argc, char ** argv) {
                        string_process_escapes(buffer);
                    }

-                    std::string user_inp = params.conversation
+                    bool format_chat = params.conversation && params.enable_chat_template;
+                    std::string user_inp = format_chat
                        ? chat_add_and_format(model, chat_msgs, "user", std::move(buffer))
                        : std::move(buffer);
                    // TODO: one inconvenient of current chat template implementation is that we can't distinguish between user input and special tokens (prefix/postfix)
                    const auto line_pfx = ::llama_tokenize(ctx, params.input_prefix, false, true);
-                    const auto line_inp = ::llama_tokenize(ctx, user_inp,            false, params.conversation);
+                    const auto line_inp = ::llama_tokenize(ctx, user_inp,            false, format_chat);
                    const auto line_sfx = ::llama_tokenize(ctx, params.input_suffix, false, true);

                    LOG("input tokens: %s\n", LOG_TOKENS_TOSTR_PRETTY(ctx, line_inp).c_str());

+                    // if user stop generation mid-way, we must add EOT to finish model's last response
+                    if (need_insert_eot && format_chat) {
+                        llama_token eot = llama_token_eot(model);
+                        embd_inp.push_back(eot == -1 ? llama_token_eos(model) : eot);
+                        need_insert_eot = false;
+                    }
+
                    embd_inp.insert(embd_inp.end(), line_pfx.begin(), line_pfx.end());
                    embd_inp.insert(embd_inp.end(), line_inp.begin(), line_inp.end());
                    embd_inp.insert(embd_inp.end(), line_sfx.begin(), line_sfx.end());
@@ -1,5 +1,8 @@
 # llama.cpp/example/passkey

+A passkey retrieval task is an evaluation method used to measure a language
+models ability to recall information from long contexts.
+
 See the following PRs for more info:

 - https://github.com/ggerganov/llama.cpp/pull/3856
@@ -1991,6 +1991,12 @@ int main(int argc, char ** argv) {
        params.n_batch = std::min(params.n_batch, n_kv);
    } else {
        params.n_batch = std::min(params.n_batch, params.n_ctx);
+        if (params.kl_divergence) {
+            params.n_parallel = 1;
+        } else {
+            // ensure there's at least enough seq_ids for HellaSwag
+            params.n_parallel = std::max(4, params.n_parallel);
+        }
    }

    if (params.ppl_stride > 0) {
@@ -2015,9 +2021,6 @@ int main(int argc, char ** argv) {
    llama_model * model;
    llama_context * ctx;

-    // ensure there's at least enough seq_ids for HellaSwag
-    params.n_parallel = std::max(4, params.n_parallel);
-
    // load the model and apply lora adapter, if any
    std::tie(model, ctx) = llama_init_from_gpt_params(params);
    if (model == NULL) {
@@ -4,7 +4,89 @@ You can also use the [GGUF-my-repo](https://huggingface.co/spaces/ggml-org/gguf-

 Note: It is synced from llama.cpp `main` every 6 hours.

-## Llama 2 7B
+Example usage:
+
+```bash
+# obtain the official LLaMA model weights and place them in ./models
+ls ./models
+llama-2-7b tokenizer_checklist.chk tokenizer.model
+# [Optional] for models using BPE tokenizers
+ls ./models
+<folder containing weights and tokenizer json> vocab.json
+# [Optional] for PyTorch .bin models like Mistral-7B
+ls ./models
+<folder containing weights and tokenizer json>
+
+# install Python dependencies
+python3 -m pip install -r requirements.txt
+
+# convert the model to ggml FP16 format
+python3 convert_hf_to_gguf.py models/mymodel/
+
+# quantize the model to 4-bits (using Q4_K_M method)
+./llama-quantize ./models/mymodel/ggml-model-f16.gguf ./models/mymodel/ggml-model-Q4_K_M.gguf Q4_K_M
+
+# update the gguf filetype to current version if older version is now unsupported
+./llama-quantize ./models/mymodel/ggml-model-Q4_K_M.gguf ./models/mymodel/ggml-model-Q4_K_M-v2.gguf COPY
+```
+
+Run the quantized model:
+
+```bash
+# start inference on a gguf model
+./llama-cli -m ./models/mymodel/ggml-model-Q4_K_M.gguf -n 128
+```
+
+When running the larger models, make sure you have enough disk space to store all the intermediate files.
+
+## Memory/Disk Requirements
+
+As the models are currently fully loaded into memory, you will need adequate disk space to save them and sufficient RAM to load them. At the moment, memory and disk requirements are the same.
+
+| Model | Original size | Quantized size (Q4_0) |
+|------:|--------------:|----------------------:|
+|    7B |         13 GB |                3.9 GB |
+|   13B |         24 GB |                7.8 GB |
+|   30B |         60 GB |               19.5 GB |
+|   65B |        120 GB |               38.5 GB |
+
+## Quantization
+
+Several quantization methods are supported. They differ in the resulting model disk size and inference speed.
+
+*(outdated)*
+
+| Model | Measure      |    F16 |   Q4_0 |   Q4_1 |   Q5_0 |   Q5_1 |   Q8_0 |
+|------:|--------------|-------:|-------:|-------:|-------:|-------:|-------:|
+|    7B | perplexity   | 5.9066 | 6.1565 | 6.0912 | 5.9862 | 5.9481 | 5.9070 |
+|    7B | file size    |  13.0G |   3.5G |   3.9G |   4.3G |   4.7G |   6.7G |
+|    7B | ms/tok @ 4th |    127 |     55 |     54 |     76 |     83 |     72 |
+|    7B | ms/tok @ 8th |    122 |     43 |     45 |     52 |     56 |     67 |
+|    7B | bits/weight  |   16.0 |    4.5 |    5.0 |    5.5 |    6.0 |    8.5 |
+|   13B | perplexity   | 5.2543 | 5.3860 | 5.3608 | 5.2856 | 5.2706 | 5.2548 |
+|   13B | file size    |  25.0G |   6.8G |   7.6G |   8.3G |   9.1G |    13G |
+|   13B | ms/tok @ 4th |      - |    103 |    105 |    148 |    160 |    131 |
+|   13B | ms/tok @ 8th |      - |     73 |     82 |     98 |    105 |    128 |
+|   13B | bits/weight  |   16.0 |    4.5 |    5.0 |    5.5 |    6.0 |    8.5 |
+
+- [k-quants](https://github.com/ggerganov/llama.cpp/pull/1684)
+- recent k-quants improvements and new i-quants
+  - [#2707](https://github.com/ggerganov/llama.cpp/pull/2707)
+  - [#2807](https://github.com/ggerganov/llama.cpp/pull/2807)
+  - [#4773 - 2-bit i-quants (inference)](https://github.com/ggerganov/llama.cpp/pull/4773)
+  - [#4856 - 2-bit i-quants (inference)](https://github.com/ggerganov/llama.cpp/pull/4856)
+  - [#4861 - importance matrix](https://github.com/ggerganov/llama.cpp/pull/4861)
+  - [#4872 - MoE models](https://github.com/ggerganov/llama.cpp/pull/4872)
+  - [#4897 - 2-bit quantization](https://github.com/ggerganov/llama.cpp/pull/4897)
+  - [#4930 - imatrix for all k-quants](https://github.com/ggerganov/llama.cpp/pull/4930)
+  - [#4951 - imatrix on the GPU](https://github.com/ggerganov/llama.cpp/pull/4957)
+  - [#4969 - imatrix for legacy quants](https://github.com/ggerganov/llama.cpp/pull/4969)
+  - [#4996 - k-qunats tuning](https://github.com/ggerganov/llama.cpp/pull/4996)
+  - [#5060 - Q3_K_XS](https://github.com/ggerganov/llama.cpp/pull/5060)
+  - [#5196 - 3-bit i-quants](https://github.com/ggerganov/llama.cpp/pull/5196)
+  - [quantization tuning](https://github.com/ggerganov/llama.cpp/pull/5320), [another one](https://github.com/ggerganov/llama.cpp/pull/5334), and [another one](https://github.com/ggerganov/llama.cpp/pull/5361)
+
+**Llama 2 7B**

 | Quantization | Bits per Weight (BPW) |
 |--------------|-----------------------|
@@ -18,7 +100,8 @@ Note: It is synced from llama.cpp `main` every 6 hours.
 | Q5_K_M       | 5.68                  |
 | Q6_K         | 6.56                  |

-## Llama 2 13B
+**Llama 2 13B**
+
 Quantization | Bits per Weight (BPW)
 -- | --
 Q2_K | 3.34
@@ -31,7 +114,7 @@ Q5_K_S | 5.51
 Q5_K_M | 5.67
 Q6_K | 6.56

-# Llama 2 70B
+**Llama 2 70B**

 Quantization | Bits per Weight (BPW)
 -- | --
@@ -73,6 +73,7 @@ The project is under active development, and we are [looking for feedback and co
 - `-fa`, `--flash-attn` : enable flash attention (default: disabled).
 - `-ctk TYPE`, `--cache-type-k TYPE` : KV cache data type for K (default: `f16`, options `f32`, `f16`, `q8_0`, `q4_0`, `q4_1`, `iq4_nl`, `q5_0`, or `q5_1`)
 - `-ctv TYPE`, `--cache-type-v TYPE` : KV cache type for V (default `f16`, see `-ctk` for options)
+- `--spm-infill` : Use Suffix/Prefix/Middle pattern for infill (instead of Prefix/Suffix/Middle) as some models prefer this.

 **If compiled with `LLAMA_SERVER_SSL=ON`**
 - `--ssl-key-file FNAME`: path to file a PEM-encoded SSL private key
@@ -365,7 +366,8 @@ Notice that each `probs` is an array of length `n_probs`.
  "assistant_name": "",
  "user_name": "",
  "default_generation_settings": { ... },
-  "total_slots": 1
+  "total_slots": 1,
+  "chat_template": ""
 }
 ```

@@ -373,8 +375,9 @@ Notice that each `probs` is an array of length `n_probs`.
 - `user_name` - the required anti-prompt to generate the prompt in case you have specified a system prompt for all slots.
 - `default_generation_settings` - the default generation settings for the `/completion` endpoint, which has the same fields as the `generation_settings` response object from the `/completion` endpoint.
 - `total_slots` - the total number of slots for process requests (defined by `--parallel` option)
+- `chat_template` - the model's original Jinja2 prompt template

- **POST** `/v1/chat/completions`: OpenAI-compatible Chat Completions API. Given a ChatML-formatted json description in `messages`, it returns the predicted completion. Both synchronous and streaming mode are supported, so scripted and interactive applications work fine. While no strong claims of compatibility with OpenAI API spec is being made, in our experience it suffices to support many apps. Only model with [supported chat template](https://github.com/ggerganov/llama.cpp/wiki/Templates-supported-by-llama_chat_apply_template) can be used optimally with this endpoint. By default, ChatML template will be used.
+- **POST** `/v1/chat/completions`: OpenAI-compatible Chat Completions API. Given a ChatML-formatted json description in `messages`, it returns the predicted completion. Both synchronous and streaming mode are supported, so scripted and interactive applications work fine. While no strong claims of compatibility with OpenAI API spec is being made, in our experience it suffices to support many apps. Only models with a [supported chat template](https://github.com/ggerganov/llama.cpp/wiki/Templates-supported-by-llama_chat_apply_template) can be used optimally with this endpoint. By default, the ChatML template will be used.

    *Options:*

@@ -259,7 +259,7 @@ const GRAMMAR_RANGE_LITERAL_ESCAPE_RE = /[\n\r"\]\-\\]/g;
 const GRAMMAR_LITERAL_ESCAPES = { '\r': '\\r', '\n': '\\n', '"': '\\"', '-': '\\-', ']': '\\]' };

 const NON_LITERAL_SET = new Set('|.()[]{}*+?');
-const ESCAPED_IN_REGEXPS_BUT_NOT_IN_LITERALS = new Set('[]()|{}*+?');
+const ESCAPED_IN_REGEXPS_BUT_NOT_IN_LITERALS = new Set('^$.[]()|{}*+?');

 export class SchemaConverter {
  constructor(options) {
@@ -751,7 +751,7 @@ export class SchemaConverter {
    const requiredProps = sortedProps.filter(k => required.has(k));
    const optionalProps = sortedProps.filter(k => !required.has(k));

-    if (additionalProperties !== false) {
+    if (additionalProperties) {
      const subName = `${name ?? ''}${name ? '-' : ''}additional`;
      const valueRule =
        additionalProperties != null && typeof additionalProperties === 'object' ? this.visit(additionalProperties, `${subName}-value`)
@@ -2020,6 +2020,7 @@ struct server_context {
                        slot.t_start_generation = 0;

                        if (slot.infill) {
+                            const bool add_bos = llama_should_add_bos_token(model);
                            bool suff_rm_leading_spc = true;
                            if (params.input_suffix.find_first_of(' ') == 0 && params.input_suffix.size() > 1) {
                                params.input_suffix.erase(0, 1);
@@ -2035,16 +2036,21 @@ struct server_context {
                            }

                            prefix_tokens.insert(prefix_tokens.begin(), llama_token_prefix(model));
-                            prefix_tokens.insert(prefix_tokens.begin(), llama_token_bos(model)); // always add BOS
-                            prefix_tokens.insert(prefix_tokens.end(),   llama_token_suffix(model));
-                            prefix_tokens.insert(prefix_tokens.end(),   suffix_tokens.begin(), suffix_tokens.end());
+                            suffix_tokens.insert(suffix_tokens.begin(), llama_token_suffix(model));
+
+                            auto embd_inp = params.spm_infill ? suffix_tokens : prefix_tokens;
+                            auto embd_end = params.spm_infill ? prefix_tokens : suffix_tokens;
+                            if (add_bos) {
+                                embd_inp.insert(embd_inp.begin(), llama_token_bos(model));
+                            }
+                            embd_inp.insert(embd_inp.end(), embd_end.begin(), embd_end.end());

                            const llama_token middle_token = llama_token_middle(model);
                            if (middle_token >= 0) {
-                                prefix_tokens.push_back(middle_token);
+                                embd_inp.push_back(middle_token);
                            }

-                            prompt_tokens = prefix_tokens;
+                            prompt_tokens = embd_inp;
                        } else {
                            prompt_tokens = tokenize(slot.prompt, system_prompt.empty()); // add BOS if there isn't system prompt
                        }
@@ -2599,7 +2605,7 @@ int main(int argc, char ** argv) {
    // if a custom chat template is not supplied, we will use the one that comes with the model (if any)
    if (params.chat_template.empty()) {
        if (!ctx_server.validate_model_chat_template()) {
-            LOG_ERROR("The chat template that comes with this model is not yet supported, falling back to chatml. This may cause the model to output suboptimal responses", {});
+            LOG_WARNING("The chat template that comes with this model is not yet supported, falling back to chatml. This may cause the model to output suboptimal responses", {});
            params.chat_template = "chatml";
        }
    }
@@ -2961,11 +2967,20 @@ int main(int argc, char ** argv) {
    };

    const auto handle_props = [&ctx_server](const httplib::Request & req, httplib::Response & res) {
+        std::string template_key = "tokenizer.chat_template", curr_tmpl;
+        int32_t tlen = llama_model_meta_val_str(ctx_server.model, template_key.c_str(), nullptr, 0);
+        if (tlen > 0) {
+            std::vector<char> curr_tmpl_buf(tlen + 1, 0);
+            if (llama_model_meta_val_str(ctx_server.model, template_key.c_str(), curr_tmpl_buf.data(), curr_tmpl_buf.size()) == tlen) {
+                curr_tmpl = std::string(curr_tmpl_buf.data(), tlen);
+            }
+        }
        res.set_header("Access-Control-Allow-Origin", req.get_header_value("Origin"));
        json data = {
            { "system_prompt",               ctx_server.system_prompt.c_str() },
            { "default_generation_settings", ctx_server.default_generation_settings_for_props },
-            { "total_slots",                 ctx_server.params.n_parallel }
+            { "total_slots",                 ctx_server.params.n_parallel },
+            { "chat_template",               curr_tmpl.c_str() }
        };

        res.set_content(data.dump(), "application/json; charset=utf-8");
@@ -52,4 +52,3 @@ Feature: Passkey / Self-extend with context shift
      #| TheBloke/Llama-2-7B-GGUF        | llama-2-7b.Q2_K.gguf        | 4096        | 3   | 16384 | 512     | 4    | 512    | 500    | 300   | 1234    | 5           | 1234           |
      #| TheBloke/Mixtral-8x7B-v0.1-GGUF | mixtral-8x7b-v0.1.Q2_K.gguf | 32768       | 2   | 16384 | 512     | 4    | 512    | 500    | 100   | 0987    | 5           | 0
      # 987           |
-
@@ -1054,4 +1054,3 @@
 </body>

 </html>
-
@@ -1058,4 +1058,3 @@
 </body>

 </html>
-
@@ -31,4 +31,3 @@ for i in range(n-1):
        embedding2 = np.array(result[j])
        similarity = np.dot(embedding1, embedding2) / (np.linalg.norm(embedding1) * np.linalg.norm(embedding2))
        print(f"Similarity between {i} and {j}: {similarity:.2f}")
-
@@ -34,4 +34,3 @@ fi

 #use multiple GPUs with same max compute units
 #ZES_ENABLE_SYSMAN=1 ./build/bin/llama-cli -m models/llama-2-7b.Q4_0.gguf -p "${INPUT2}" -n 400 -e -ngl 33 -s 0
-
@@ -31,4 +31,3 @@ exit /B 0
 :ERROR
 echo comomand error: %errorlevel%
 exit /B %errorlevel%
-
@@ -7,5 +7,3 @@ set INPUT2="Building a website can be done in 10 simple steps:\nStep 1:"


 .\build\bin\main.exe -m models\llama-2-7b.Q4_0.gguf -p %INPUT2% -n 400 -e -ngl 33 -s 0
-
-
@@ -30,6 +30,7 @@ static void print_usage_information(const char * argv0, FILE * stream) {
    fprintf(stream, "    --stdin                              read prompt from standard input.\n");
    fprintf(stream, "    --no-bos                             do not ever add a BOS token to the prompt, even if normally the model uses a BOS token.\n");
    fprintf(stream, "    --log-disable                        disable logs. Makes stderr quiet when loading the model.\n");
+    fprintf(stream, "    --show-count                         print the total number of tokens.\n");
 }

 static void llama_log_callback_null(ggml_log_level level, const char * text, void * user_data) {
@@ -195,6 +196,7 @@ int main(int raw_argc, char ** raw_argv) {
    bool printing_ids = false;
    bool no_bos = false;
    bool disable_logging = false;
+    bool show_token_count = false;
    const char * model_path = NULL;
    const char * prompt_path = NULL;
    const char * prompt_arg = NULL;
@@ -249,6 +251,9 @@ int main(int raw_argc, char ** raw_argv) {
        else if (arg == "--log-disable") {
            disable_logging = true;
        }
+        else if (arg == "--show-count") {
+            show_token_count = true;
+        }
        else {
            fprintf(stderr, "Error: unknown option '%s'\n", argv[iarg].c_str());
            return 1;
@@ -384,6 +389,9 @@ int main(int raw_argc, char ** raw_argv) {
        printf("]\n");
    }

+    if (show_token_count) {
+        printf("Total number of tokens: %ld\n", tokens.size());
+    }
    // silence valgrind
    llama_free(ctx);
    llama_free_model(model);
@@ -20,11 +20,11 @@
    },
    "nixpkgs": {
      "locked": {
-        "lastModified": 1718318537,
-        "narHash": "sha256-4Zu0RYRcAY/VWuu6awwq4opuiD//ahpc2aFHg2CWqFY=",
+        "lastModified": 1719506693,
+        "narHash": "sha256-C8e9S7RzshSdHB7L+v9I51af1gDM5unhJ2xO1ywxNH8=",
        "owner": "NixOS",
        "repo": "nixpkgs",
-        "rev": "e9ee548d90ff586a6471b4ae80ae9cfcbceb3420",
+        "rev": "b2852eb9365c6de48ffb0dc2c9562591f652242a",
        "type": "github"
      },
      "original": {
@@ -109,6 +109,7 @@ option(GGML_LLAMAFILE                       "ggml: use ggml SGEMM"
 option(GGML_CUDA                            "ggml: use CUDA"                                  OFF)
 option(GGML_CUDA_FORCE_DMMV                 "ggml: use dmmv instead of mmvq CUDA kernels"     OFF)
 option(GGML_CUDA_FORCE_MMQ                  "ggml: use mmq kernels instead of cuBLAS"         OFF)
+option(GGML_CUDA_FORCE_CUBLAS               "ggml: always use cuBLAS instead of mmq kernels"  OFF)
 set   (GGML_CUDA_DMMV_X   "32" CACHE STRING "ggml: x stride for dmmv CUDA kernels")
 set   (GGML_CUDA_MMV_Y     "1" CACHE STRING "ggml: y block size for mmv CUDA kernels")
 option(GGML_CUDA_F16                        "ggml: use 16 bit floats for some calculations"   OFF)
@@ -119,6 +120,7 @@ set   (GGML_CUDA_PEER_MAX_BATCH_SIZE "128" CACHE STRING
 option(GGML_CUDA_NO_PEER_COPY               "ggml: do not use peer to peer copies"            OFF)
 option(GGML_CUDA_NO_VMM                     "ggml: do not try to use CUDA VMM"                OFF)
 option(GGML_CUDA_FA_ALL_QUANTS              "ggml: compile all quants for FlashAttention"     OFF)
+option(GGML_CUDA_USE_GRAPHS                 "ggml: use CUDA graphs (llama.cpp only)"          OFF)

 option(GGML_CURL                            "ggml: use libcurl to download model from an URL" OFF)
 option(GGML_HIPBLAS                         "ggml: use hipBLAS"                               OFF)
@@ -63,4 +63,3 @@ GGML_API void ggml_backend_metal_capture_next_compute(ggml_backend_t backend);
 #ifdef __cplusplus
 }
 #endif
-
@@ -295,12 +295,15 @@ if (GGML_CUDA)

        list(APPEND GGML_CDEF_PUBLIC GGML_USE_CUDA)

-        add_compile_definitions(GGML_CUDA_USE_GRAPHS)
        add_compile_definitions(GGML_CUDA_DMMV_X=${GGML_CUDA_DMMV_X})
        add_compile_definitions(GGML_CUDA_MMV_Y=${GGML_CUDA_MMV_Y})
        add_compile_definitions(K_QUANTS_PER_ITERATION=${GGML_CUDA_KQUANTS_ITER})
        add_compile_definitions(GGML_CUDA_PEER_MAX_BATCH_SIZE=${GGML_CUDA_PEER_MAX_BATCH_SIZE})

+        if (GGML_CUDA_USE_GRAPHS)
+            add_compile_definitions(GGML_CUDA_USE_GRAPHS)
+        endif()
+
        if (GGML_CUDA_FORCE_DMMV)
            add_compile_definitions(GGML_CUDA_FORCE_DMMV)
        endif()
@@ -483,9 +486,11 @@ if (GGML_SYCL)
    add_compile_options(-I./) #include DPCT

    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-narrowing")
-    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -O3")
    if (GGML_SYCL_TARGET STREQUAL "NVIDIA")
        set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fsycl-targets=nvptx64-nvidia-cuda")
+        add_compile_definitions(GGML_SYCL_WARP_SIZE=32)
+    else()
+        add_compile_definitions(GGML_SYCL_WARP_SIZE=16)
    endif()

    file(GLOB   GGML_HEADERS_SYCL "ggml-sycl/*.hpp")
@@ -1163,9 +1168,12 @@ target_link_libraries(ggml PRIVATE Threads::Threads ${GGML_EXTRA_LIBS})

 find_library(MATH_LIBRARY m)
 if (MATH_LIBRARY)
-    target_link_libraries(ggml PRIVATE ${MATH_LIBRARY})
+    if (NOT WIN32 OR NOT GGML_SYCL)
+        target_link_libraries(ggml PRIVATE ${MATH_LIBRARY})
+    endif()
 endif()

 if (BUILD_SHARED_LIBS)
    set_target_properties(ggml PROPERTIES POSITION_INDEPENDENT_CODE ON)
+    target_compile_definitions(ggml PRIVATE GGML_SHARED GGML_BUILD)
 endif()
@@ -106,19 +106,19 @@ typedef sycl::half2 ggml_half2;
 #define QR6_K 2

 #define QI2_XXS (QK_K / (4*QR2_XXS))
-#define QR2_XXS 8
+#define QR2_XXS 4

 #define QI2_XS (QK_K / (4*QR2_XS))
-#define QR2_XS 8
+#define QR2_XS 4

 #define QI2_S (QK_K / (4*QR2_S))
-#define QR2_S 8
+#define QR2_S 4

 #define QI3_XXS (QK_K / (4*QR3_XXS))
-#define QR3_XXS 8
+#define QR3_XXS 4

 #define QI3_XS (QK_K / (4*QR3_XS))
-#define QR3_XS 8
+#define QR3_XS 4

 #define QI1_S (QK_K / (4*QR1_S))
 #define QR1_S 8
@@ -130,10 +130,10 @@ typedef sycl::half2 ggml_half2;
 #define QR4_NL 2

 #define QI4_XS (QK_K / (4*QR4_XS))
-#define QR4_XS 8
+#define QR4_XS 2

 #define QI3_S (QK_K / (4*QR3_S))
-#define QR3_S 8
+#define QR3_S 4

 #endif // GGML_COMMON_DECL_CUDA || GGML_COMMON_DECL_HIP

@@ -1882,6 +1882,11 @@ static void ggml_cuda_mul_mat(ggml_backend_cuda_context & ctx, const ggml_tensor
    bool              use_mul_mat_q =  ggml_is_quantized(src0->type)
        && src1->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_F32;

+    // if mmvq is available it's a better choice than dmmv:
+#ifndef GGML_CUDA_FORCE_DMMV
+    use_dequantize_mul_mat_vec = use_dequantize_mul_mat_vec && !use_mul_mat_vec_q;
+#endif // GGML_CUDA_FORCE_DMMV
+
    bool any_gpus_with_slow_fp16 = false;

    if (split) {
@@ -1894,22 +1899,15 @@ static void ggml_cuda_mul_mat(ggml_backend_cuda_context & ctx, const ggml_tensor
            }

            const int cc            = ggml_cuda_info().devices[id].cc;
-            use_mul_mat_vec_q       = use_mul_mat_vec_q       && cc >= MIN_CC_DP4A;
            use_mul_mat_q           = use_mul_mat_q           && ggml_cuda_should_use_mmq(src0->type, cc, src1->ne[1]);
            any_gpus_with_slow_fp16 = any_gpus_with_slow_fp16 || !fast_fp16_available(cc);
        }
    } else {
        const int cc            = ggml_cuda_info().devices[ctx.device].cc;
-        use_mul_mat_vec_q       = use_mul_mat_vec_q       && cc >= MIN_CC_DP4A;
        use_mul_mat_q           = use_mul_mat_q           && ggml_cuda_should_use_mmq(src0->type, cc, src1->ne[1]);
        any_gpus_with_slow_fp16 = any_gpus_with_slow_fp16 || !fast_fp16_available(cc);
    }

-    // if mmvq is available it's a better choice than dmmv:
-#ifndef GGML_CUDA_FORCE_DMMV
-    use_dequantize_mul_mat_vec = use_dequantize_mul_mat_vec && !use_mul_mat_vec_q;
-#endif // GGML_CUDA_FORCE_DMMV
-
    // debug helpers
    //printf("src0: %8d %8d %8d %8d\n", src0->ne[0], src0->ne[1], src0->ne[2], src0->ne[3]);
    //printf("      %8d %8d %8d %8d\n", src0->nb[0], src0->nb[1], src0->nb[2], src0->nb[3]);
@@ -2713,27 +2711,40 @@ GGML_CALL static bool ggml_backend_cuda_supports_op(ggml_backend_t backend, cons
        case GGML_OP_MUL_MAT:
        case GGML_OP_MUL_MAT_ID:
            {
-                struct ggml_tensor * a;
-                struct ggml_tensor * b;
+                struct ggml_tensor * a = op->src[0];
                if (op->op == GGML_OP_MUL_MAT) {
-                    a = op->src[0];
-                    b = op->src[1];
-                } else {
-                    a = op->src[2];
-                    b = op->src[1];
-                }
-                if (a->ne[3] != b->ne[3]) {
-                    return false;
-                }
-                ggml_type a_type = a->type;
-                if (a_type == GGML_TYPE_IQ2_XXS || a_type == GGML_TYPE_IQ2_XS || a_type == GGML_TYPE_IQ3_XXS ||
-                    a_type == GGML_TYPE_IQ1_S   || a_type == GGML_TYPE_IQ4_NL || a_type == GGML_TYPE_IQ3_S   ||
-                    a_type == GGML_TYPE_IQ1_M   || a_type == GGML_TYPE_IQ2_S  || a_type == GGML_TYPE_IQ4_XS) {
-                    if (b->ne[1] == 1 && ggml_nrows(b) > 1) {
+                    struct ggml_tensor * b = op->src[1];
+                    if (a->ne[3] != b->ne[3]) {
                        return false;
                    }
                }
-                return true;
+                switch (a->type) {
+                    case GGML_TYPE_F32:
+                    case GGML_TYPE_F16:
+                    case GGML_TYPE_Q4_0:
+                    case GGML_TYPE_Q4_1:
+                    case GGML_TYPE_Q5_0:
+                    case GGML_TYPE_Q5_1:
+                    case GGML_TYPE_Q8_0:
+                    case GGML_TYPE_Q2_K:
+                    case GGML_TYPE_Q3_K:
+                    case GGML_TYPE_Q4_K:
+                    case GGML_TYPE_Q5_K:
+                    case GGML_TYPE_Q6_K:
+                    case GGML_TYPE_Q8_K:
+                    case GGML_TYPE_IQ1_M:
+                    case GGML_TYPE_IQ1_S:
+                    case GGML_TYPE_IQ2_S:
+                    case GGML_TYPE_IQ2_XS:
+                    case GGML_TYPE_IQ2_XXS:
+                    case GGML_TYPE_IQ3_S:
+                    case GGML_TYPE_IQ3_XXS:
+                    case GGML_TYPE_IQ4_NL:
+                    case GGML_TYPE_IQ4_XS:
+                        return true;
+                    default:
+                        return false;
+                }
            } break;
        case GGML_OP_GET_ROWS:
            {
@@ -3,6 +3,7 @@
 #include "ggml.h"
 #include "ggml-cuda.h"

+#include <cstdint>
 #include <memory>

 #if defined(GGML_USE_HIPBLAS)
@@ -226,6 +227,10 @@ typedef float2 dfloat2;
 #define RDNA2
 #endif

+#if defined(__gfx1010__) || defined(__gfx1012__)
+#define RDNA1
+#endif
+
 #ifndef __has_builtin
    #define __has_builtin(x) 0
 #endif
@@ -268,30 +273,15 @@ static __device__ __forceinline__ unsigned int __vcmpeq4(unsigned int a, unsigne
    return c;
 }

-static __device__ __forceinline__ int __dp4a(const int a, const int b, int c) {
-#if defined(__gfx906__) || defined(__gfx908__) || defined(__gfx90a__) || defined(__gfx1030__)
-    c = __builtin_amdgcn_sdot4(a, b, c, false);
-#elif defined(RDNA3)
-    c = __builtin_amdgcn_sudot4( true, a, true, b, c, false);
-#elif defined(__gfx1010__) || defined(__gfx900__)
-    int tmp1;
-    int tmp2;
-    asm("\n \
-        v_mul_i32_i24 %1, sext(%3), sext(%4) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:BYTE_0 \n \
-        v_mul_i32_i24 %2, sext(%3), sext(%4) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1 \n \
-        v_add3_u32 %0, %1, %2, %0 \n \
-        v_mul_i32_i24 %1, sext(%3), sext(%4) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_2 src1_sel:BYTE_2 \n \
-        v_mul_i32_i24 %2, sext(%3), sext(%4) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:BYTE_3 \n \
-        v_add3_u32 %0, %1, %2, %0 \n \
-        "
-        : "+v"(c), "=&v"(tmp1), "=&v"(tmp2)
-        : "v"(a), "v"(b)
-    );
-#else
-    const int8x4_t va = reinterpret_cast<const int8x4_t&>(a);
-    const int8x4_t vb = reinterpret_cast<const int8x4_t&>(b);
-    c += va[0] * vb[0] + va[1] * vb[1] + va[2] * vb[2] + va[3] * vb[3];
-#endif
+static __device__ __forceinline__ unsigned int __vcmpne4(unsigned int a, unsigned int b) {
+    const uint8x4_t& va = reinterpret_cast<const uint8x4_t&>(a);
+    const uint8x4_t& vb = reinterpret_cast<const uint8x4_t&>(b);
+    unsigned int c;
+    uint8x4_t& vc = reinterpret_cast<uint8x4_t&>(c);
+#pragma unroll
+    for (int i = 0; i < 4; ++i) {
+        vc[i] = va[i] == vb[i] ? 0x00 : 0xff;
+    }
    return c;
 }

@@ -467,8 +457,48 @@ static __device__ __forceinline__ uint32_t __hgt2_mask(const half2 a, const half
 }
 #endif // CUDART_VERSION < 12000

+static __device__ __forceinline__ int ggml_cuda_dp4a(const int a, const int b, int c) {
+#if defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__)
+#if defined(__gfx906__) || defined(__gfx908__) || defined(__gfx90a__) || defined(__gfx1030__)
+    c = __builtin_amdgcn_sdot4(a, b, c, false);
+#elif defined(RDNA3)
+    c = __builtin_amdgcn_sudot4( true, a, true, b, c, false);
+#elif defined(__gfx1010__) || defined(__gfx900__)
+    int tmp1;
+    int tmp2;
+    asm("\n \
+        v_mul_i32_i24 %1, sext(%3), sext(%4) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:BYTE_0 \n \
+        v_mul_i32_i24 %2, sext(%3), sext(%4) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1 \n \
+        v_add3_u32 %0, %1, %2, %0 \n \
+        v_mul_i32_i24 %1, sext(%3), sext(%4) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_2 src1_sel:BYTE_2 \n \
+        v_mul_i32_i24 %2, sext(%3), sext(%4) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:BYTE_3 \n \
+        v_add3_u32 %0, %1, %2, %0 \n \
+        "
+        : "+v"(c), "=&v"(tmp1), "=&v"(tmp2)
+        : "v"(a), "v"(b)
+    );
+#else
+    const int8x4_t va = reinterpret_cast<const int8x4_t&>(a);
+    const int8x4_t vb = reinterpret_cast<const int8x4_t&>(b);
+    c += va[0] * vb[0] + va[1] * vb[1] + va[2] * vb[2] + va[3] * vb[3];
+#endif
+    return c;
+
+#else // defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__)
+
+#if __CUDA_ARCH__ >= MIN_CC_DP4A
+    return __dp4a(a, b, c);
+#else // __CUDA_ARCH__ >= MIN_CC_DP4A
+    const int8_t * a8 = (const int8_t *) &a;
+    const int8_t * b8 = (const int8_t *) &b;
+    return c + a8[0]*b8[0] + a8[1]*b8[1] + a8[2]*b8[2] + a8[3]*b8[3];
+#endif // __CUDA_ARCH__ >= MIN_CC_DP4A
+
+#endif // defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__)
+}
+
 // TODO: move to ggml-common.h
-static const __device__ int8_t kvalues_iq4nl[16] = {-127, -104, -83, -65, -49, -35, -22, -10, 1, 13, 25, 38, 53, 69, 89, 113};
+static constexpr __device__ int8_t kvalues_iq4nl[16] = {-127, -104, -83, -65, -49, -35, -22, -10, 1, 13, 25, 38, 53, 69, 89, 113};

 typedef void (*dequantize_kernel_t)(const void * vx, const int64_t ib, const int iqs, dfloat2 & v);

@@ -487,4 +487,3 @@ void* ggml_cuda_cpy_fn(const ggml_tensor * src0, ggml_tensor * src1) {
        GGML_ASSERT(false);
    }
 }
-
@@ -54,12 +54,11 @@ typedef float (*vec_dot_KQ_f32_t)(
 template<typename T, int D>
 static __device__ __forceinline__ T vec_dot_fattn_vec_KQ_q4_0(
    const char * __restrict__ K_c, const void * __restrict__ Q_v, const int * __restrict__ Q_q8, const void * __restrict__ Q_ds_v) {
-#if __CUDA_ARCH__ >= MIN_CC_DP4A

    const block_q4_0 * K_q4_0 = (const block_q4_0 *) K_c;
    GGML_UNUSED(Q_v);

-    half sum = 0.0f;
+    T sum = 0.0f;

 #pragma unroll
    for (int k_KQ_0 = 0; k_KQ_0 < D/sizeof(int); k_KQ_0 += WARP_SIZE) {
@@ -69,10 +68,10 @@ static __device__ __forceinline__ T vec_dot_fattn_vec_KQ_q4_0(
        const int iqs4  = k_KQ %  QI4_0;
        const int shift = k_KQ & (QI8_1/2);

-        const int v = (get_int_from_uint8(K_q4_0[ib].qs, iqs4) >> shift) & 0x0F0F0F0F;
+        const int v = (get_int_b2(K_q4_0[ib].qs, iqs4) >> shift) & 0x0F0F0F0F;
        const int u = Q_q8[k_KQ_0/WARP_SIZE];

-        const int sumi = __dp4a(v, u, 0);
+        const int sumi = ggml_cuda_dp4a(v, u, 0);

 #ifdef FP16_AVAILABLE
        if (std::is_same<T, half>::value) {
@@ -90,19 +89,11 @@ static __device__ __forceinline__ T vec_dot_fattn_vec_KQ_q4_0(
    }

    return sum;
-#else
-    GGML_UNUSED(K_c);
-    GGML_UNUSED(Q_v);
-    GGML_UNUSED(Q_q8);
-    GGML_UNUSED(Q_ds_v);
-    NO_DEVICE_CODE;
-#endif  // __CUDA_ARCH__ >= MIN_CC_DP4A
 }

 template<typename T, int D>
 static __device__ __forceinline__ T vec_dot_fattn_vec_KQ_q4_1(
    const char * __restrict__ K_c, const void * __restrict__ Q_v, const int * __restrict__ Q_q8, const void * __restrict__ Q_ds_v) {
-#if __CUDA_ARCH__ >= MIN_CC_DP4A

    const block_q4_1 * K_q4_1 = (const block_q4_1 *) K_c;
    GGML_UNUSED(Q_v);
@@ -117,10 +108,10 @@ static __device__ __forceinline__ T vec_dot_fattn_vec_KQ_q4_1(
        const int iqs4  = k_KQ %  QI4_1;
        const int shift = k_KQ & (QI8_1/2);

-        const int v = (get_int_from_uint8_aligned(K_q4_1[ib].qs, iqs4) >> shift) & 0x0F0F0F0F;
+        const int v = (get_int_b4(K_q4_1[ib].qs, iqs4) >> shift) & 0x0F0F0F0F;
        const int u = Q_q8[k_KQ_0/WARP_SIZE];

-        const int sumi = __dp4a(v, u, 0);
+        const int sumi = ggml_cuda_dp4a(v, u, 0);

 #ifdef FP16_AVAILABLE
        if (std::is_same<T, half>::value) {
@@ -142,19 +133,11 @@ static __device__ __forceinline__ T vec_dot_fattn_vec_KQ_q4_1(
    }

    return sum;
-#else
-    GGML_UNUSED(K_c);
-    GGML_UNUSED(Q_v);
-    GGML_UNUSED(Q_q8);
-    GGML_UNUSED(Q_ds_v);
-    NO_DEVICE_CODE;
-#endif  // __CUDA_ARCH__ >= MIN_CC_DP4A
 }

 template<typename T, int D>
 static __device__ __forceinline__ T vec_dot_fattn_vec_KQ_q5_0(
    const char * __restrict__ K_c, const void * __restrict__ Q_v, const int * __restrict__ Q_q8, const void * __restrict__ Q_ds_v) {
-#if __CUDA_ARCH__ >= MIN_CC_DP4A

    const block_q5_0 * K_q5_0 = (const block_q5_0 *) K_c;
    GGML_UNUSED(Q_v);
@@ -170,8 +153,8 @@ static __device__ __forceinline__ T vec_dot_fattn_vec_KQ_q5_0(
        const int iqs8  = k_KQ %  QI8_1;
        const int shift = k_KQ & (QI8_1/2);

-        int v = (get_int_from_uint8(K_q5_0[ib].qs, iqs4) >> shift) & 0x0F0F0F0F;
-        const int vh = get_int_from_uint8(K_q5_0[ib].qh, 0) >> (iqs8 * QI5_0);
+        int v = (get_int_b2(K_q5_0[ib].qs, iqs4) >> shift) & 0x0F0F0F0F;
+        const int vh = get_int_b2(K_q5_0[ib].qh, 0) >> (iqs8 * QI5_0);
        v |= (vh <<  4) & 0x00000010; // 0 ->  4
        v |= (vh << 11) & 0x00001000; // 1 -> 12
        v |= (vh << 18) & 0x00100000; // 2 -> 20
@@ -179,7 +162,7 @@ static __device__ __forceinline__ T vec_dot_fattn_vec_KQ_q5_0(

        const int u = Q_q8[k_KQ_0/WARP_SIZE];

-        const int sumi = __dp4a(v, u, 0);
+        const int sumi = ggml_cuda_dp4a(v, u, 0);

 #ifdef FP16_AVAILABLE
        if (std::is_same<T, half>::value) {
@@ -197,19 +180,11 @@ static __device__ __forceinline__ T vec_dot_fattn_vec_KQ_q5_0(
    }

    return sum;
-#else
-    GGML_UNUSED(K_c);
-    GGML_UNUSED(Q_v);
-    GGML_UNUSED(Q_q8);
-    GGML_UNUSED(Q_ds_v);
-    NO_DEVICE_CODE;
-#endif  // __CUDA_ARCH__ >= MIN_CC_DP4A
 }

 template<typename T, int D>
 static __device__ __forceinline__ T vec_dot_fattn_vec_KQ_q5_1(
    const char * __restrict__ K_c, const void * __restrict__ Q_v, const int * __restrict__ Q_q8, const void * __restrict__ Q_ds_v) {
-#if __CUDA_ARCH__ >= MIN_CC_DP4A

    const block_q5_1 * K_q5_1 = (const block_q5_1 *) K_c;
    GGML_UNUSED(Q_v);
@@ -225,8 +200,8 @@ static __device__ __forceinline__ T vec_dot_fattn_vec_KQ_q5_1(
        const int iqs8  = k_KQ %  QI8_1;
        const int shift = k_KQ & (QI8_1/2);

-        int v = (get_int_from_uint8(K_q5_1[ib].qs, iqs4) >> shift) & 0x0F0F0F0F;
-        const int vh = get_int_from_uint8(K_q5_1[ib].qh, 0) >> (iqs8 * QI5_1);
+        int v = (get_int_b2(K_q5_1[ib].qs, iqs4) >> shift) & 0x0F0F0F0F;
+        const int vh = get_int_b2(K_q5_1[ib].qh, 0) >> (iqs8 * QI5_1);
        v |= (vh <<  4) & 0x00000010; // 0 ->  4
        v |= (vh << 11) & 0x00001000; // 1 -> 12
        v |= (vh << 18) & 0x00100000; // 2 -> 20
@@ -234,7 +209,7 @@ static __device__ __forceinline__ T vec_dot_fattn_vec_KQ_q5_1(

        const int u = Q_q8[k_KQ_0/WARP_SIZE];

-        const int sumi = __dp4a(v, u, 0);
+        const int sumi = ggml_cuda_dp4a(v, u, 0);

 #ifdef FP16_AVAILABLE
        if (std::is_same<T, half>::value) {
@@ -256,19 +231,11 @@ static __device__ __forceinline__ T vec_dot_fattn_vec_KQ_q5_1(
    }

    return sum;
-#else
-    GGML_UNUSED(K_c);
-    GGML_UNUSED(Q_v);
-    GGML_UNUSED(Q_q8);
-    GGML_UNUSED(Q_ds_v);
-    NO_DEVICE_CODE;
-#endif  // __CUDA_ARCH__ >= MIN_CC_DP4A
 }

 template <typename T, int D>
 static __device__ __forceinline__ T vec_dot_fattn_vec_KQ_q8_0(
    const char * __restrict__ K_c, const void * __restrict__ Q_v, const int * __restrict__ Q_q8, const void * __restrict__ Q_ds_v) {
-#if __CUDA_ARCH__ >= MIN_CC_DP4A

    const block_q8_0 * K_q8_0 = (const block_q8_0 *) K_c;
    GGML_UNUSED(Q_v);
@@ -282,7 +249,7 @@ static __device__ __forceinline__ T vec_dot_fattn_vec_KQ_q8_0(
        const int ib  = k_KQ / QI8_0;
        const int iqs = k_KQ % QI8_0;

-        const int v = get_int_from_int8(K_q8_0[ib].qs, iqs);
+        const int v = get_int_b2(K_q8_0[ib].qs, iqs);

        T Q_d;
        if (std::is_same<T, half>::value) {
@@ -297,13 +264,6 @@ static __device__ __forceinline__ T vec_dot_fattn_vec_KQ_q8_0(
    }

    return sum;
-#else
-    GGML_UNUSED(K_c);
-    GGML_UNUSED(Q_v);
-    GGML_UNUSED(Q_q8);
-    GGML_UNUSED(Q_ds_v);
-    NO_DEVICE_CODE;
-#endif  // __CUDA_ARCH__ >= MIN_CC_DP4A
 }

 template <typename T, int D>
@@ -448,7 +408,7 @@ static __device__ __forceinline__ T dequantize_1_q5_0(const void * __restrict__

    const T   d   = x[ib].d;
    const int ql0 = x[ib].qs[iqs];
-    const int qh0 = get_int_from_uint8(x[ib].qh, 0);
+    const int qh0 = get_int_b2(x[ib].qh, 0);
    const int ql  = ((ql0 >> (4*shift)) & 0x0F);
    const int qh  = ((qh0 >> idq) << 4) & 0x10;
    const int q   = (ql | qh) - 16;
@@ -473,7 +433,7 @@ static __device__ __forceinline__ T dequantize_1_q5_1(const void * __restrict__

    const half2 dm  = x[ib].dm;
    const int   ql0 = x[ib].qs[iqs];
-    const int   qh0 = get_int_from_uint8_aligned(x[ib].qh, 0);
+    const int   qh0 = get_int_b4(x[ib].qh, 0);
    const int   ql  = ((ql0 >> (4*shift)) & 0x0F);
    const int   qh  = ((qh0 >> idq) << 4) & 0x10;
    const int   q   = (ql | qh);
@@ -59,6 +59,12 @@ void ggml_cuda_op_mul_mat_q(
        case GGML_TYPE_Q6_K:
            mul_mat_q_case<GGML_TYPE_Q6_K>(ctx, args, stream);
            break;
+        case GGML_TYPE_IQ4_XS:
+            mul_mat_q_case<GGML_TYPE_IQ4_XS>(ctx, args, stream);
+            break;
+        case GGML_TYPE_IQ4_NL:
+            mul_mat_q_case<GGML_TYPE_IQ4_NL>(ctx, args, stream);
+            break;
        default:
            GGML_ASSERT(false);
            break;
@@ -87,6 +93,8 @@ bool ggml_cuda_should_use_mmq(enum ggml_type type, int cc, int64_t ne11) {
        case GGML_TYPE_Q4_K:
        case GGML_TYPE_Q5_K:
        case GGML_TYPE_Q6_K:
+        case GGML_TYPE_IQ4_XS:
+        case GGML_TYPE_IQ4_NL:
            mmq_supported = true;
            break;
        default:
@@ -60,12 +60,16 @@ static constexpr __device__ int get_mmq_x_max_device() {
 }

 static constexpr int get_mmq_y_host(const int cc) {
-    return int8_mma_available(cc) || cc >= CC_VOLTA ? 128 : 64;
+    return cc >= CC_OFFSET_AMD ? (cc == CC_RDNA1 ? 64 : 128) : (cc >= CC_VOLTA ? 128 : 64);
 }

 static constexpr __device__ int get_mmq_y_device() {
 #if defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__)
+#if defined(RDNA1)
+    return 64;
+#else
    return 128;
+#endif // defined RDNA1
 #else
 #if __CUDA_ARCH__ >= CC_VOLTA
    return 128;
@@ -88,15 +92,17 @@ static constexpr __device__ int get_mmq_y_device() {

 static constexpr __host__ __device__ tile_x_sizes mmq_get_dp4a_tile_x_sizes(ggml_type type, int mmq_y) {
    return type == GGML_TYPE_Q4_0 ? MMQ_DP4A_TXS_Q4_0 :
-        type == GGML_TYPE_Q4_1 ? MMQ_DP4A_TXS_Q4_1 :
-        type == GGML_TYPE_Q5_0 ? MMQ_DP4A_TXS_Q5_0 :
-        type == GGML_TYPE_Q5_1 ? MMQ_DP4A_TXS_Q5_1 :
-        type == GGML_TYPE_Q8_0 ? MMQ_DP4A_TXS_Q8_0 :
-        type == GGML_TYPE_Q2_K ? MMQ_DP4A_TXS_Q2_K :
-        type == GGML_TYPE_Q3_K ? MMQ_DP4A_TXS_Q3_K :
-        type == GGML_TYPE_Q4_K ? MMQ_DP4A_TXS_Q4_K :
-        type == GGML_TYPE_Q5_K ? MMQ_DP4A_TXS_Q5_K :
-        type == GGML_TYPE_Q6_K ? MMQ_DP4A_TXS_Q6_K :
+        type == GGML_TYPE_Q4_1    ? MMQ_DP4A_TXS_Q4_1 :
+        type == GGML_TYPE_Q5_0    ? MMQ_DP4A_TXS_Q5_0 :
+        type == GGML_TYPE_Q5_1    ? MMQ_DP4A_TXS_Q5_1 :
+        type == GGML_TYPE_Q8_0    ? MMQ_DP4A_TXS_Q8_0 :
+        type == GGML_TYPE_Q2_K    ? MMQ_DP4A_TXS_Q2_K :
+        type == GGML_TYPE_Q3_K    ? MMQ_DP4A_TXS_Q3_K :
+        type == GGML_TYPE_Q4_K    ? MMQ_DP4A_TXS_Q4_K :
+        type == GGML_TYPE_Q5_K    ? MMQ_DP4A_TXS_Q5_K :
+        type == GGML_TYPE_Q6_K    ? MMQ_DP4A_TXS_Q6_K :
+        type == GGML_TYPE_IQ4_XS  ? MMQ_DP4A_TXS_Q5_0 :
+        type == GGML_TYPE_IQ4_NL  ? MMQ_DP4A_TXS_Q5_0 :
        tile_x_sizes{0, 0, 0};
 }

@@ -124,15 +130,17 @@ static_assert(MMQ_MMA_TILE_X_K_Q6_K % 8 == 4, "Wrong padding.");

 static constexpr __host__ __device__ int mmq_get_mma_tile_x_k(ggml_type type) {
    return type == GGML_TYPE_Q4_0 ? MMQ_MMA_TILE_X_K_Q4_0 :
-        type == GGML_TYPE_Q4_1 ? MMQ_MMA_TILE_X_K_Q4_1 :
-        type == GGML_TYPE_Q5_0 ? MMQ_MMA_TILE_X_K_Q5_0 :
-        type == GGML_TYPE_Q5_1 ? MMQ_MMA_TILE_X_K_Q5_1 :
-        type == GGML_TYPE_Q8_0 ? MMQ_MMA_TILE_X_K_Q8_0 :
-        type == GGML_TYPE_Q2_K ? MMQ_MMA_TILE_X_K_Q2_K :
-        type == GGML_TYPE_Q3_K ? MMQ_MMA_TILE_X_K_Q3_K :
-        type == GGML_TYPE_Q4_K ? MMQ_MMA_TILE_X_K_Q4_K :
-        type == GGML_TYPE_Q5_K ? MMQ_MMA_TILE_X_K_Q5_K :
-        type == GGML_TYPE_Q6_K ? MMQ_MMA_TILE_X_K_Q6_K :
+        type == GGML_TYPE_Q4_1    ? MMQ_MMA_TILE_X_K_Q4_1 :
+        type == GGML_TYPE_Q5_0    ? MMQ_MMA_TILE_X_K_Q5_0 :
+        type == GGML_TYPE_Q5_1    ? MMQ_MMA_TILE_X_K_Q5_1 :
+        type == GGML_TYPE_Q8_0    ? MMQ_MMA_TILE_X_K_Q8_0 :
+        type == GGML_TYPE_Q2_K    ? MMQ_MMA_TILE_X_K_Q2_K :
+        type == GGML_TYPE_Q3_K    ? MMQ_MMA_TILE_X_K_Q3_K :
+        type == GGML_TYPE_Q4_K    ? MMQ_MMA_TILE_X_K_Q4_K :
+        type == GGML_TYPE_Q5_K    ? MMQ_MMA_TILE_X_K_Q5_K :
+        type == GGML_TYPE_Q6_K    ? MMQ_MMA_TILE_X_K_Q6_K :
+        type == GGML_TYPE_IQ4_XS  ? MMQ_MMA_TILE_X_K_Q5_0 :
+        type == GGML_TYPE_IQ4_NL  ? MMQ_MMA_TILE_X_K_Q5_0 :
        0;
 }

@@ -181,9 +189,9 @@ template <int mmq_y, int nwarps, bool need_check> static __device__ __forceinlin
        const block_q4_0 * bxi = (const block_q4_0 *) x + kbx0 + i*stride + kbx;

 #ifdef INT8_MMA_AVAILABLE
-        x_qs[i*MMQ_MMA_TILE_X_K_Q4_0 + threadIdx.x] = get_int_from_uint8(bxi->qs, kqsx);
+        x_qs[i*MMQ_MMA_TILE_X_K_Q4_0 + threadIdx.x] = get_int_b2(bxi->qs, kqsx);
 #else
-        x_qs[i*(WARP_SIZE + 1)       + threadIdx.x] = get_int_from_uint8(bxi->qs, kqsx);
+        x_qs[i*(WARP_SIZE + 1)       + threadIdx.x] = get_int_b2(bxi->qs, kqsx);
 #endif // INT8_MMA_AVAILABLE
    }

@@ -344,9 +352,9 @@ template <int mmq_y, int nwarps, bool need_check> static __device__ __forceinlin
        const block_q4_1 * bxi = (const block_q4_1 *) x + kbx0 + i*stride + kbx;

 #ifdef INT8_MMA_AVAILABLE
-        x_qs[i*MMQ_MMA_TILE_X_K_Q4_1 + threadIdx.x] = get_int_from_uint8_aligned(bxi->qs, kqsx);
+        x_qs[i*MMQ_MMA_TILE_X_K_Q4_1 + threadIdx.x] = get_int_b4(bxi->qs, kqsx);
 #else
-        x_qs[i*(WARP_SIZE + 1)       + threadIdx.x] = get_int_from_uint8_aligned(bxi->qs, kqsx);
+        x_qs[i*(WARP_SIZE + 1)       + threadIdx.x] = get_int_b4(bxi->qs, kqsx);
 #endif // INT8_MMA_AVAILABLE
    }

@@ -505,8 +513,8 @@ template <int mmq_y, int nwarps, bool need_check> static __device__ __forceinlin

        const block_q5_0 * bxi = (const block_q5_0 *) x + kbx0 + i*stride + kbx;

-        const int ql = get_int_from_uint8(bxi->qs, kqsx);
-        const int qh = get_int_from_uint8(bxi->qh, 0) >> (4 * (threadIdx.x % QI5_0));
+        const int ql = get_int_b2(bxi->qs, kqsx);
+        const int qh = get_int_b2(bxi->qh, 0) >> (4 * (threadIdx.x % QI5_0));

        int qs0 = (ql >>  0)   & 0x0F0F0F0F;
        qs0    |= (qh <<  4)   & 0x00000010;  // 0 ->  4
@@ -670,8 +678,8 @@ template <int mmq_y, int nwarps, bool need_check> static __device__ __forceinlin

        const block_q5_1 * bxi = (const block_q5_1 *) x + kbx0 + i*stride + kbx;

-        const int ql = get_int_from_uint8_aligned(bxi->qs, kqsx);
-        const int qh = get_int_from_uint8_aligned(bxi->qh, 0) >> (4 * (threadIdx.x % QI5_1));
+        const int ql = get_int_b4(bxi->qs, kqsx);
+        const int qh = get_int_b4(bxi->qh, 0) >> (4 * (threadIdx.x % QI5_1));

        int qs0 = (ql >>  0) & 0x0F0F0F0F;
        qs0    |= (qh <<  4) & 0x00000010; // 0 ->  4
@@ -835,9 +843,9 @@ template <int mmq_y, int nwarps, bool need_check> static __device__ __forceinlin
        const block_q8_0 * bxi = (const block_q8_0 *) x + kbx0 + i*stride + kbx;

 #ifdef INT8_MMA_AVAILABLE
-        x_qs[i*MMQ_MMA_TILE_X_K_Q8_0 + threadIdx.x] = get_int_from_int8(bxi->qs, kqsx);
+        x_qs[i*MMQ_MMA_TILE_X_K_Q8_0 + threadIdx.x] = get_int_b2(bxi->qs, kqsx);
 #else
-        x_qs[i*(WARP_SIZE + 1)       + threadIdx.x] = get_int_from_int8(bxi->qs, kqsx);
+        x_qs[i*(WARP_SIZE + 1)       + threadIdx.x] = get_int_b2(bxi->qs, kqsx);
 #endif // INT8_MMA_AVAILABLE
    }

@@ -980,7 +988,7 @@ template <int mmq_y, int nwarps, bool need_check> static __device__ __forceinlin

        const block_q2_K * bxi = (const block_q2_K *) x + kbx0 + i*stride + kbx;

-        const int x_ql_0 = get_int_from_uint8(bxi->qs, kqsx);
+        const int x_ql_0 = get_int_b2(bxi->qs, kqsx);

 #pragma unroll
        for (int l = 0; l < QR2_K; ++l) {
@@ -1162,8 +1170,8 @@ template <int mmq_y, int nwarps, bool need_check> static __device__ __forceinlin

        const block_q3_K * bxi = (const block_q3_K *) x + kbx0 + i*stride + kbx;

-        const int x_ql_0 = get_int_from_uint8(bxi->qs,    kqsx);
-        const int x_qh_0 = get_int_from_uint8(bxi->hmask, kqsx % (QI3_K/2)) >> (4 * (kqsx / (QI3_K/2)));
+        const int x_ql_0 = get_int_b2(bxi->qs,    kqsx);
+        const int x_qh_0 = get_int_b2(bxi->hmask, kqsx % (QI3_K/2)) >> (4 * (kqsx / (QI3_K/2)));

 #pragma unroll
        for (int l = 0; l < QR3_K; ++l) {
@@ -1221,11 +1229,11 @@ template <int mmq_y, int nwarps, bool need_check> static __device__ __forceinlin

        const int ksc_low = ksc % (QI3_K/8);
        const int shift_low = 4 * (ksc / (QI3_K/8));
-        const int sc_low = (get_int_from_uint8(bxi->scales, ksc_low) >> shift_low) & 0x0F0F0F0F;
+        const int sc_low = (get_int_b2(bxi->scales, ksc_low) >> shift_low) & 0x0F0F0F0F;

        const int ksc_high = QI3_K/8;
        const int shift_high = 2 * ksc;
-        const int sc_high = ((get_int_from_uint8(bxi->scales, ksc_high) >> shift_high) << 4) & 0x30303030;
+        const int sc_high = ((get_int_b2(bxi->scales, ksc_high) >> shift_high) << 4) & 0x30303030;

        const int sc = __vsubss4(sc_low | sc_high, 0x20202020);

@@ -1389,9 +1397,9 @@ template <int mmq_y, int nwarps, bool need_check> static __device__ __forceinlin
        const block_q4_K * bxi = (const block_q4_K *) x + kbx0 + i*stride + kbx;

 #ifdef INT8_MMA_AVAILABLE
-        x_qs[i*MMQ_MMA_TILE_X_K_Q4_K + threadIdx.x] = get_int_from_uint8_aligned(bxi->qs, kqsx);
+        x_qs[i*MMQ_MMA_TILE_X_K_Q4_K + threadIdx.x] = get_int_b4(bxi->qs, kqsx);
 #else
-        x_qs[i*(WARP_SIZE + 1)       + threadIdx.x] = get_int_from_uint8_aligned(bxi->qs, kqsx);
+        x_qs[i*(WARP_SIZE + 1)       + threadIdx.x] = get_int_b4(bxi->qs, kqsx);
 #endif // INT8_MMA_AVAILABLE
    }

@@ -1606,11 +1614,11 @@ template <int mmq_y, int nwarps, bool need_check> static __device__ __forceinlin
        const block_q5_K * bxi = (const block_q5_K *) x + kbx0 + i*stride + kbx;
        const int ky = QR5_K*kqsx;

-        const int ql = get_int_from_uint8_aligned(bxi->qs, kqsx);
+        const int ql = get_int_b4(bxi->qs, kqsx);
        const int ql0 = (ql >> 0) & 0x0F0F0F0F;
        const int ql1 = (ql >> 4) & 0x0F0F0F0F;

-        const int qh = get_int_from_uint8_aligned(bxi->qh, kqsx % (QI5_K/4));
+        const int qh = get_int_b4(bxi->qh, kqsx % (QI5_K/4));
        const int qh0 = ((qh >> (2 * (kqsx / (QI5_K/4)) + 0)) << 4) & 0x10101010;
        const int qh1 = ((qh >> (2 * (kqsx / (QI5_K/4)) + 1)) << 4) & 0x10101010;

@@ -1828,11 +1836,11 @@ template <int mmq_y, int nwarps, bool need_check> static __device__ __forceinlin
        const block_q6_K * bxi = (const block_q6_K *) x + kbx0 + i*stride + kbx;
        const int ky = QR6_K*kqsx;

-        const int ql = get_int_from_uint8(bxi->ql, kqsx);
+        const int ql = get_int_b2(bxi->ql, kqsx);
        const int ql0 = (ql >> 0) & 0x0F0F0F0F;
        const int ql1 = (ql >> 4) & 0x0F0F0F0F;

-        const int qh = get_int_from_uint8(bxi->qh, (QI6_K/4) * (kqsx / (QI6_K/2)) + kqsx % (QI6_K/4));
+        const int qh = get_int_b2(bxi->qh, (QI6_K/4) * (kqsx / (QI6_K/2)) + kqsx % (QI6_K/4));
        const int qh0 = ((qh >> (2 * ((kqsx % (QI6_K/2)) / (QI6_K/4)))) << 4) & 0x30303030;
        const int qh1 =  (qh >> (2 * ((kqsx % (QI6_K/2)) / (QI6_K/4))))       & 0x30303030;

@@ -1879,9 +1887,9 @@ template <int mmq_y, int nwarps, bool need_check> static __device__ __forceinlin
        const block_q6_K * bxi = (const block_q6_K *) x + kbx0 + i*stride + (threadIdx.x % (WARP_SIZE/8)) / 4;

 #ifdef INT8_MMA_AVAILABLE
-        x_sc[i*MMQ_MMA_TILE_X_K_Q6_K + threadIdx.x % (WARP_SIZE/8)] = get_int_from_int8(bxi->scales, threadIdx.x % (QI6_K/8));
+        x_sc[i*MMQ_MMA_TILE_X_K_Q6_K + threadIdx.x % (WARP_SIZE/8)] = get_int_b2(bxi->scales, threadIdx.x % (QI6_K/8));
 #else
-        x_sc[i*(WARP_SIZE/8) + i/8   + threadIdx.x % (WARP_SIZE/8)] = get_int_from_int8(bxi->scales, threadIdx.x % (QI6_K/8));
+        x_sc[i*(WARP_SIZE/8) + i/8   + threadIdx.x % (WARP_SIZE/8)] = get_int_b2(bxi->scales, threadIdx.x % (QI6_K/8));
 #endif // INT8_MMA_AVAILABLE
    }
 }
@@ -2014,6 +2022,124 @@ static __device__ __forceinline__ void vec_dot_q6_K_q8_1_mma(
 #endif // INT8_MMA_AVAILABLE
 }

+template <int mmq_y, int nwarps, bool need_check> static __device__ __forceinline__ void load_tiles_iq4_nl(
+    const char * __restrict__ x, int * __restrict__ x_tile, const int & kbx0, const int & i_max, const int & stride) {
+
+#ifdef INT8_MMA_AVAILABLE
+    int   * x_qs = (int   *)  x_tile;
+    float * x_df = (float *) (x_qs + WARP_SIZE*2);
+#else
+    constexpr tile_x_sizes txs = mmq_get_dp4a_tile_x_sizes(GGML_TYPE_IQ4_NL, mmq_y);
+    int   * x_qs = (int   *)  x_tile;
+    float * x_df = (float *) (x_qs + txs.qs);
+#endif // INT8_MMA_AVAILABLE
+
+    const int kbx  = threadIdx.x / QI4_NL;
+    const int kqsx = threadIdx.x % QI4_NL;
+
+#pragma unroll
+    for (int i0 = 0; i0 < mmq_y; i0 += nwarps) {
+        int i = i0 + threadIdx.y;
+
+        if (need_check) {
+            i = min(i, i_max);
+        }
+
+        const block_iq4_nl * bxi = (const block_iq4_nl *) x + kbx0 + i*stride + kbx;
+
+        const int aux_q4 = get_int_b2(bxi->qs, kqsx);
+        const int2 v = get_int_from_table_16(aux_q4);
+        const int k0 = 8 * (threadIdx.x / 4) + threadIdx.x % 4;
+#ifdef INT8_MMA_AVAILABLE
+        x_qs[i*MMQ_MMA_TILE_X_K_Q5_0 + k0 + 0] = v.x;
+        x_qs[i*MMQ_MMA_TILE_X_K_Q5_0 + k0 + 4] = v.y;
+#else
+        x_qs[i*(2*WARP_SIZE + 1)     + k0 + 0] = v.x;
+        x_qs[i*(2*WARP_SIZE + 1)     + k0 + 4] = v.y;
+#endif // INT8_MMA_AVAILABLE
+    }
+
+    const int blocks_per_tile_x_row = WARP_SIZE / QI4_NL;
+    const int kbxd = threadIdx.x % blocks_per_tile_x_row;
+
+#pragma unroll
+    for (int i0 = 0; i0 < mmq_y; i0 += nwarps * QI4_NL) {
+        int i = i0 + threadIdx.y * QI4_NL + threadIdx.x / blocks_per_tile_x_row;
+
+        if (need_check) {
+            i = min(i, i_max);
+        }
+
+        const block_iq4_nl * bxi = (const block_iq4_nl *) x + kbx0 + i*stride + kbxd;
+
+#ifdef INT8_MMA_AVAILABLE
+        x_df[i*MMQ_MMA_TILE_X_K_Q5_0 + kbxd] = __half2float(bxi->d);
+#else
+        x_df[i*(WARP_SIZE/4) + i/4   + kbxd] = __half2float(bxi->d);
+#endif // INT8_MMA_AVAILABLE
+    }
+}
+
+template <int mmq_y, int nwarps, bool need_check> static __device__ __forceinline__ void load_tiles_iq4_xs(
+    const char * __restrict__ x, int * __restrict__ x_tile, const int & kbx0, const int & i_max, const int & stride) {
+
+#ifdef INT8_MMA_AVAILABLE
+    int   * x_qs = (int   *)  x_tile;
+    float * x_df = (float *) (x_qs + WARP_SIZE*2);
+#else
+    constexpr tile_x_sizes txs = mmq_get_dp4a_tile_x_sizes(GGML_TYPE_IQ4_XS, mmq_y);
+    int   * x_qs = (int   *)  x_tile;
+    float * x_df = (float *) (x_qs + txs.qs);
+#endif // INT8_MMA_AVAILABLE
+
+    const int kbx  = 0;           // threadIdx.x / QI4_XS
+    const int kqsx = threadIdx.x; // threadIdx.x % QI4_XS
+
+#pragma unroll
+    for (int i0 = 0; i0 < mmq_y; i0 += nwarps) {
+        int i = i0 + threadIdx.y;
+
+        if (need_check) {
+            i = min(i, i_max);
+        }
+
+        const block_iq4_xs * bxi = (const block_iq4_xs *) x + kbx0 + i*stride + kbx;
+
+        const int aux_q4 = get_int_b4(bxi->qs, kqsx);
+        const int2 v = get_int_from_table_16(aux_q4);
+        const int k0 = 8 * (threadIdx.x / 4) + threadIdx.x % 4;
+#ifdef INT8_MMA_AVAILABLE
+        x_qs[i*MMQ_MMA_TILE_X_K_Q5_0 + k0 + 0] = v.x;
+        x_qs[i*MMQ_MMA_TILE_X_K_Q5_0 + k0 + 4] = v.y;
+#else
+        x_qs[i*(2*WARP_SIZE + 1)     + k0 + 0] = v.x;
+        x_qs[i*(2*WARP_SIZE + 1)     + k0 + 4] = v.y;
+#endif // INT8_MMA_AVAILABLE
+    }
+
+#pragma unroll
+    for (int i0 = 0; i0 < mmq_y; i0 += nwarps * 4) {
+        int i = i0 + threadIdx.y * 4 + threadIdx.x / (WARP_SIZE/4);
+
+        if (need_check) {
+            i = min(i, i_max);
+        }
+
+        const block_iq4_xs * bxi = (const block_iq4_xs *) x + kbx0 + i*stride;
+
+        const float d = __half2float(bxi->d);
+
+        const int ls = ((bxi->scales_l[(threadIdx.x % 8)/2] >> (4*(threadIdx.x % 2))) & 0x0F)
+            | (((bxi->scales_h >> (2*(threadIdx.x % 8))) & 0x03) << 4);
+
+#ifdef INT8_MMA_AVAILABLE
+        x_df[i*MMQ_MMA_TILE_X_K_Q5_0 + threadIdx.x % 8] = d * (ls - 32);
+#else
+        x_df[i*(WARP_SIZE/4) + i/4   + threadIdx.x % 8] = d * (ls - 32);
+#endif // INT8_MMA_AVAILABLE
+    }
+}
+
 template<int mmq_x, int mmq_y, int nwarps, bool need_check>
 static __device__ __forceinline__ void mmq_write_back_dp4a(
    const float * __restrict__ sum, float * __restrict__ dst, const int & stride, const int & i_max, const int & j_max) {
@@ -2163,6 +2289,22 @@ struct mmq_type_traits<mmq_x, mmq_y, nwarps, need_check, GGML_TYPE_Q6_K> {
    static constexpr vec_dot_mmq_t    vec_dot_dp4a = vec_dot_q6_K_q8_1_dp4a<mmq_x, mmq_y, nwarps>;
 };

+template <int mmq_x, int mmq_y, int nwarps, bool need_check>
+struct mmq_type_traits<mmq_x, mmq_y, nwarps, need_check, GGML_TYPE_IQ4_NL> {
+    static constexpr int              vdr          = VDR_IQ4_NL_Q8_1_MMQ;
+    static constexpr load_tiles_mmq_t load_tiles   = load_tiles_iq4_nl<mmq_y, nwarps, need_check>;
+    static constexpr vec_dot_mmq_t    vec_dot_mma  = vec_dot_q5_0_q8_1_mma<mmq_x, mmq_y, nwarps>;
+    static constexpr vec_dot_mmq_t    vec_dot_dp4a = vec_dot_q5_0_q8_1_dp4a<mmq_x, mmq_y, nwarps>;
+};
+
+template <int mmq_x, int mmq_y, int nwarps, bool need_check>
+struct mmq_type_traits<mmq_x, mmq_y, nwarps, need_check, GGML_TYPE_IQ4_XS> {
+    static constexpr int              vdr          = VDR_IQ4_XS_Q8_1_MMQ;
+    static constexpr load_tiles_mmq_t load_tiles   = load_tiles_iq4_xs<mmq_y, nwarps, need_check>;
+    static constexpr vec_dot_mmq_t    vec_dot_mma  = vec_dot_q5_0_q8_1_mma<mmq_x, mmq_y, nwarps>;
+    static constexpr vec_dot_mmq_t    vec_dot_dp4a = vec_dot_q5_0_q8_1_dp4a<mmq_x, mmq_y, nwarps>;
+};
+
 static bool mmq_need_sum(const ggml_type type_x) {
    switch (type_x) {
        case GGML_TYPE_Q4_0:
@@ -2180,6 +2322,8 @@ static bool mmq_need_sum(const ggml_type type_x) {
        case GGML_TYPE_Q5_K:
            return true;
        case GGML_TYPE_Q6_K:
+        case GGML_TYPE_IQ4_XS:
+        case GGML_TYPE_IQ4_NL:
            return false;
        default:
            GGML_ASSERT(false);
@@ -2301,8 +2445,11 @@ static __global__ void mul_mat_q(
    const int nty = (ne01 + mmq_y - 1) / mmq_y; // Number of tiles y

    // kbc == k block continuous, current index in continuous ijk space.
-    int64_t       kbc      = GGML_PAD((int64_t) blockIdx.x     *blocks_per_ne00*ntx*nty / gridDim.x, blocks_per_warp);
-    const int64_t kbc_stop = GGML_PAD((int64_t)(blockIdx.x + 1)*blocks_per_ne00*ntx*nty / gridDim.x, blocks_per_warp);
+    int64_t kbc      = (int64_t) blockIdx.x     *blocks_per_ne00*ntx*nty / gridDim.x;
+    int64_t kbc_stop = (int64_t)(blockIdx.x + 1)*blocks_per_ne00*ntx*nty / gridDim.x;
+
+    kbc      -= (kbc      % blocks_per_ne00) % blocks_per_warp;
+    kbc_stop -= (kbc_stop % blocks_per_ne00) % blocks_per_warp;

    // kb0 == k index when doing the matrix multiplication for an output tile.
    int kb0_start = kbc % blocks_per_ne00;
@@ -2358,8 +2505,11 @@ static __global__ void mul_mat_q_stream_k_fixup(
    const int bidx_stop  = (blockIdx.y*nty + blockIdx.x + 1) * block_num_mmq / (gridDim.y*gridDim.x) + 1;

    for (int bidx = bidx_start; bidx < bidx_stop; ++bidx) {
-        const int64_t kbc      = GGML_PAD((int64_t) bidx     *blocks_per_ne00*ntx*nty / block_num_mmq, blocks_per_warp);
-        const int64_t kbc_stop = GGML_PAD((int64_t)(bidx + 1)*blocks_per_ne00*ntx*nty / block_num_mmq, blocks_per_warp);
+        int64_t kbc      = (int64_t) bidx     *blocks_per_ne00*ntx*nty / block_num_mmq;
+        int64_t kbc_stop = (int64_t)(bidx + 1)*blocks_per_ne00*ntx*nty / block_num_mmq;
+
+        kbc      -= (kbc      % blocks_per_ne00) % blocks_per_warp;
+        kbc_stop -= (kbc_stop % blocks_per_ne00) % blocks_per_warp;

        // Skip fixup tile if the MMQ CUDA block never wrote anything to it:
        if (kbc == kbc_stop || kbc_stop % blocks_per_ne00 == 0) {
@@ -2475,7 +2625,7 @@ static void launch_mul_mat_q(ggml_backend_cuda_context & ctx, const mmq_args & a

    const dim3 block_nums_mmq(nsm, 1, 1);

-    ggml_cuda_pool & pool = ctx.pool();
+    ggml_cuda_pool & pool = ctx.pool(id);
    ggml_cuda_pool_alloc<float> tmp_fixup(pool, block_nums_mmq.x * mmq_x*mmq_y);

    if (args.ne01 % mmq_y == 0) {
@@ -2598,6 +2748,8 @@ extern DECL_MMQ_CASE(GGML_TYPE_Q3_K);
 extern DECL_MMQ_CASE(GGML_TYPE_Q4_K);
 extern DECL_MMQ_CASE(GGML_TYPE_Q5_K);
 extern DECL_MMQ_CASE(GGML_TYPE_Q6_K);
+extern DECL_MMQ_CASE(GGML_TYPE_IQ4_NL);
+extern DECL_MMQ_CASE(GGML_TYPE_IQ4_XS);

 // -------------------------------------------------------------------------------------------------------------------------

@@ -28,16 +28,22 @@ static constexpr __device__ vec_dot_q_cuda_t get_vec_dot_q_cuda(ggml_type type)

 static constexpr __device__ int get_vdr_mmvq(ggml_type type) {
    return type == GGML_TYPE_Q4_0 ? VDR_Q4_0_Q8_1_MMVQ :
-        type == GGML_TYPE_Q4_1 ? VDR_Q4_1_Q8_1_MMVQ :
-        type == GGML_TYPE_Q5_0 ? VDR_Q5_0_Q8_1_MMVQ :
-        type == GGML_TYPE_Q5_1 ? VDR_Q5_1_Q8_1_MMVQ :
-        type == GGML_TYPE_Q8_0 ? VDR_Q8_0_Q8_1_MMVQ :
-        type == GGML_TYPE_Q2_K ? VDR_Q2_K_Q8_1_MMVQ :
-        type == GGML_TYPE_Q3_K ? VDR_Q3_K_Q8_1_MMVQ :
-        type == GGML_TYPE_Q4_K ? VDR_Q4_K_Q8_1_MMVQ :
-        type == GGML_TYPE_Q5_K ? VDR_Q5_K_Q8_1_MMVQ :
-        type == GGML_TYPE_Q6_K ? VDR_Q6_K_Q8_1_MMVQ :
-        type == GGML_TYPE_IQ4_NL ? VDR_Q4_K_Q8_1_MMVQ :
+        type == GGML_TYPE_Q4_1    ? VDR_Q4_1_Q8_1_MMVQ :
+        type == GGML_TYPE_Q5_0    ? VDR_Q5_0_Q8_1_MMVQ :
+        type == GGML_TYPE_Q5_1    ? VDR_Q5_1_Q8_1_MMVQ :
+        type == GGML_TYPE_Q8_0    ? VDR_Q8_0_Q8_1_MMVQ :
+        type == GGML_TYPE_Q2_K    ? VDR_Q2_K_Q8_1_MMVQ :
+        type == GGML_TYPE_Q3_K    ? VDR_Q3_K_Q8_1_MMVQ :
+        type == GGML_TYPE_Q4_K    ? VDR_Q4_K_Q8_1_MMVQ :
+        type == GGML_TYPE_Q5_K    ? VDR_Q5_K_Q8_1_MMVQ :
+        type == GGML_TYPE_Q6_K    ? VDR_Q6_K_Q8_1_MMVQ :
+        type == GGML_TYPE_IQ2_XXS ? VDR_IQ2_XXS_Q8_1_MMVQ :
+        type == GGML_TYPE_IQ2_XS  ? VDR_IQ2_XS_Q8_1_MMVQ :
+        type == GGML_TYPE_IQ2_S   ? VDR_IQ2_S_Q8_1_MMVQ :
+        type == GGML_TYPE_IQ3_XXS ? VDR_IQ3_XXS_Q8_1_MMVQ :
+        type == GGML_TYPE_IQ3_S   ? VDR_IQ3_S_Q8_1_MMVQ :
+        type == GGML_TYPE_IQ4_NL  ? VDR_IQ4_NL_Q8_1_MMVQ :
+        type == GGML_TYPE_IQ4_XS  ? VDR_IQ4_XS_Q8_1_MMVQ :
        1;
 }

@@ -22,7 +22,8 @@ SOURCE_FATTN_WMMA_CASE = "DECL_FATTN_WMMA_F16_CASE({head_size}, {cols_per_block}

 TYPES_MMQ = [
    "GGML_TYPE_Q4_0", "GGML_TYPE_Q4_1", "GGML_TYPE_Q5_0", "GGML_TYPE_Q5_1", "GGML_TYPE_Q8_0",
-    "GGML_TYPE_Q2_K", "GGML_TYPE_Q3_K", "GGML_TYPE_Q4_K", "GGML_TYPE_Q5_K", "GGML_TYPE_Q6_K"
+    "GGML_TYPE_Q2_K", "GGML_TYPE_Q3_K", "GGML_TYPE_Q4_K", "GGML_TYPE_Q5_K", "GGML_TYPE_Q6_K",
+    "GGML_TYPE_IQ4_NL", "GGML_TYPE_IQ4_XS"
 ]

 SOURCE_MMQ = """// This file has been autogenerated by generate_cu_files.py, do not edit manually.
@@ -0,0 +1,5 @@
+// This file has been autogenerated by generate_cu_files.py, do not edit manually.
+
+#include "../mmq.cuh"
+
+DECL_MMQ_CASE(GGML_TYPE_IQ4_NL);
@@ -0,0 +1,5 @@
+// This file has been autogenerated by generate_cu_files.py, do not edit manually.
+
+#include "../mmq.cuh"
+
+DECL_MMQ_CASE(GGML_TYPE_IQ4_XS);
--- a/Show More
+++ b/Show More
				`@@ -7,5 +7,3 @@ set INPUT2="Building a website can be done in 10 simple steps:\nStep 1:"`


				`.\build\bin\main.exe -m models\llama-2-7b.Q4_0.gguf -p %INPUT2% -n 400 -e -ngl 33 -s 0`