server: fix incorrectly reported token probabilities (#7125)

* server: normalize token probabilities

* fix temperature == 0.0f

CMakeLists.txt

@@ -103,6 +103,8 @@ set(LLAMA_CUDA_KQUANTS_ITER "2" CACHE STRING "llama: iters./thread per block for
 set(LLAMA_CUDA_PEER_MAX_BATCH_SIZE "128" CACHE STRING
                                              "llama: max. batch size for using peer access")
 option(LLAMA_CUDA_NO_PEER_COPY               "llama: do not use peer to peer copies"            OFF)
+option(LLAMA_CUDA_NO_VMM                     "llama: do not try to use CUDA VMM"                OFF)
+
 option(LLAMA_CURL                            "llama: use libcurl to download model from an URL" OFF)
 option(LLAMA_HIPBLAS                         "llama: use hipBLAS"                               OFF)
 option(LLAMA_HIP_UMA                         "llama: use HIP unified memory architecture"       OFF)
@@ -409,6 +411,9 @@ if (LLAMA_CUDA)
         if (LLAMA_CUDA_FORCE_MMQ)
             add_compile_definitions(GGML_CUDA_FORCE_MMQ)
         endif()
+        if (LLAMA_CUDA_NO_VMM)
+            add_compile_definitions(GGML_CUDA_NO_VMM)
+        endif()
         add_compile_definitions(GGML_CUDA_DMMV_X=${LLAMA_CUDA_DMMV_X})
         add_compile_definitions(GGML_CUDA_MMV_Y=${LLAMA_CUDA_MMV_Y})
         if (DEFINED LLAMA_CUDA_DMMV_Y)
@@ -434,7 +439,11 @@ if (LLAMA_CUDA)
             set(LLAMA_EXTRA_LIBS ${LLAMA_EXTRA_LIBS} CUDA::cudart CUDA::cublas CUDA::cublasLt)
         endif()
 
-        set(LLAMA_EXTRA_LIBS ${LLAMA_EXTRA_LIBS} CUDA::cuda_driver)
+        if (LLAMA_CUDA_NO_VMM)
+            # No VMM requested, no need to link directly with the cuda driver lib (libcuda.so)
+        else()
+            set(LLAMA_EXTRA_LIBS ${LLAMA_EXTRA_LIBS} CUDA::cuda_driver) # required by cuDeviceGetAttribute(), cuMemGetAllocationGranularity(...), ...
+        endif()
 
     if (NOT DEFINED CMAKE_CUDA_ARCHITECTURES)
         # 52 == lowest CUDA 12 standard

README.md

@@ -20,7 +20,8 @@ Inference of Meta's [LLaMA](https://arxiv.org/abs/2302.13971) model (and others)
 
 ### Hot topics
 
-- **BPE pre-tokenization support has been added: https://github.com/ggerganov/llama.cpp/pull/6920**
+- **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
 - Model sharding instructions using `gguf-split` https://github.com/ggerganov/llama.cpp/discussions/6404
 - Fix major bug in Metal batched inference https://github.com/ggerganov/llama.cpp/pull/6225
@@ -712,7 +713,7 @@ Building the program with BLAS support may lead to some performance improvements
 
 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` does not support LLaMA 3, you can use `convert-hf-to-gguf.py` with LLaMA 3 downloaded from Hugging Face. 
+Note: `convert.py` does not support LLaMA 3, you can use `convert-hf-to-gguf.py` with LLaMA 3 downloaded from Hugging Face.
 
 ```bash
 # obtain the official LLaMA model weights and place them in ./models

ci/run.sh

@@ -160,9 +160,8 @@ function gg_run_test_scripts_debug {
 
     set -e
 
-    # TODO: too slow, run on dedicated node
-   #(cd ./examples/gguf-split && time bash tests.sh "$SRC/build-ci-debug/bin" "$MNT/models") 2>&1 | tee -a $OUT/${ci}-scripts.log
-   #(cd ./examples/quantize   && time bash tests.sh "$SRC/build-ci-debug/bin" "$MNT/models") 2>&1 | tee -a $OUT/${ci}-scripts.log
+    (cd ./examples/gguf-split && time bash tests.sh "$SRC/build-ci-debug/bin" "$MNT/models") 2>&1 | tee -a $OUT/${ci}-scripts.log
+    (cd ./examples/quantize   && time bash tests.sh "$SRC/build-ci-debug/bin" "$MNT/models") 2>&1 | tee -a $OUT/${ci}-scripts.log
 
     set +e
 }
@@ -695,8 +694,10 @@ test $ret -eq 0 && gg_run ctest_release
 if [ -z ${GG_BUILD_LOW_PERF} ]; then
     test $ret -eq 0 && gg_run embd_bge_small
 
-    test $ret -eq 0 && gg_run test_scripts_debug
-    test $ret -eq 0 && gg_run test_scripts_release
+    if [ -z ${GG_BUILD_CLOUD} ] || [ ${GG_BUILD_EXTRA_TESTS_0} ]; then
+        test $ret -eq 0 && gg_run test_scripts_debug
+        test $ret -eq 0 && gg_run test_scripts_release
+    fi
 
     if [ -z ${GG_BUILD_VRAM_GB} ] || [ ${GG_BUILD_VRAM_GB} -ge 8 ]; then
         if [ -z ${GG_BUILD_CUDA} ]; then

common/sampling.cpp

@@ -35,6 +35,8 @@ struct llama_sampling_context * llama_sampling_init(const struct llama_sampling_
 
     result->prev.resize(params.n_prev);
 
+    result->n_considered = 0;
+
     llama_sampling_set_rng_seed(result, params.seed);
 
     return result;
@@ -64,6 +66,7 @@ void llama_sampling_reset(llama_sampling_context * ctx) {
 
     std::fill(ctx->prev.begin(), ctx->prev.end(), 0);
     ctx->cur.clear();
+    ctx->n_considered = 0;
 }
 
 void llama_sampling_set_rng_seed(struct llama_sampling_context * ctx, uint32_t seed) {
@@ -253,6 +256,8 @@ static llama_token llama_sampling_sample_impl(
         }
     }
 
+    ctx_sampling->n_considered = cur_p.size;
+
     return id;
 }
 

common/sampling.h

@@ -81,6 +81,7 @@ struct llama_sampling_context {
     // TODO: replace with ring-buffer
     std::vector<llama_token>      prev;
     std::vector<llama_token_data> cur;
+    size_t n_considered;
 
     std::mt19937 rng;
 };

convert-hf-to-gguf-update.py

@@ -67,6 +67,7 @@ models = [
     {"name": "gpt-2",          "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/openai-community/gpt2", },
     {"name": "refact",         "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/smallcloudai/Refact-1_6-base", },
     {"name": "command-r",      "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/CohereForAI/c4ai-command-r-v01", },
+    {"name": "olmo",           "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/allenai/OLMo-1.7-7B-hf", },
 ]
 
 # make directory "models/tokenizers" if it doesn't exist

convert-hf-to-gguf.py

@@ -314,6 +314,9 @@ class Model(ABC):
         if chkhsh == "9c2227e4dd922002fb81bde4fc02b0483ca4f12911410dee2255e4987644e3f8":
             # ref: https://huggingface.co/CohereForAI/c4ai-command-r-v01
             res = "command-r"
+        if chkhsh == "b6dc8df998e1cfbdc4eac8243701a65afe638679230920b50d6f17d81c098166":
+            # ref: https://huggingface.co/allenai/OLMo-1.7-7B-hf
+            res = "olmo"
 
         if res is None:
             logger.warning("\n")
@@ -2831,8 +2834,9 @@ class OlmoModel(Model):
     def set_gguf_parameters(self):
         super().set_gguf_parameters()
         self.gguf_writer.add_layer_norm_eps(1e-5)
-        if "clip_qkv" in self.hparams is not None:
-            self.gguf_writer.add_clamp_kqv(self.hparams["clip_qkv"])
+        clip_qkv = self.hparams.get("clip_qkv")
+        if clip_qkv is not None:
+            self.gguf_writer.add_clamp_kqv(clip_qkv)
 
     # Same as super class, but permuting q_proj, k_proj
     # Copied from: LlamaModel

docs/BLIS.md

@@ -23,7 +23,7 @@ Install BLIS:
 sudo make install
 ```
 
-We recommend using openmp since it's easier to modify the cores been used.
+We recommend using openmp since it's easier to modify the cores being used.
 
 ### llama.cpp compilation
 

docs/HOWTO-add-model.md

@@ -96,9 +96,9 @@ NOTE: The dimensions in `ggml` are typically in the reverse order of the `pytorc
 
 This is the funniest part, you have to provide the inference graph implementation of the new model architecture in `llama_build_graph`.
 
-Have a look to existing implementation like `build_llama`, `build_dbrx` or `build_bert`.
+Have a look at existing implementation like `build_llama`, `build_dbrx` or `build_bert`.
 
-When implementing a new graph, please note that the underlying `ggml` backends might not support them all, support of missing backend operations can be added in another PR.
+When implementing a new graph, please note that the underlying `ggml` backends might not support them all, support for missing backend operations can be added in another PR.
 
 Note: to debug the inference graph: you can use [eval-callback](../examples/eval-callback).
 

examples/llama-bench/llama-bench.cpp

@@ -178,6 +178,7 @@ struct cmd_params {
     std::vector<std::vector<float>> tensor_split;
     std::vector<bool> use_mmap;
     std::vector<bool> embeddings;
+    ggml_numa_strategy numa;
     int reps;
     bool verbose;
     output_formats output_format;
@@ -200,6 +201,7 @@ static const cmd_params cmd_params_defaults = {
     /* tensor_split  */ {std::vector<float>(llama_max_devices(), 0.0f)},
     /* use_mmap      */ {true},
     /* embeddings    */ {false},
+    /* numa          */ GGML_NUMA_STRATEGY_DISABLED,
     /* reps          */ 5,
     /* verbose       */ false,
     /* output_format */ MARKDOWN
@@ -224,6 +226,7 @@ static void print_usage(int /* argc */, char ** argv) {
     printf("  -nkvo, --no-kv-offload <0|1>        (default: %s)\n", join(cmd_params_defaults.no_kv_offload, ",").c_str());
     printf("  -fa, --flash-attn <0|1>             (default: %s)\n", join(cmd_params_defaults.flash_attn, ",").c_str());
     printf("  -mmp, --mmap <0|1>                  (default: %s)\n", join(cmd_params_defaults.use_mmap, ",").c_str());
+    printf("  --numa <distribute|isolate|numactl> (default: disabled)\n");
     printf("  -embd, --embeddings <0|1>           (default: %s)\n", join(cmd_params_defaults.embeddings, ",").c_str());
     printf("  -ts, --tensor-split <ts0/ts1/..>    (default: 0)\n");
     printf("  -r, --repetitions <n>               (default: %d)\n", cmd_params_defaults.reps);
@@ -396,6 +399,17 @@ static cmd_params parse_cmd_params(int argc, char ** argv) {
             }
             auto p = split<bool>(argv[i], split_delim);
             params.no_kv_offload.insert(params.no_kv_offload.end(), p.begin(), p.end());
+        } else if (arg == "--numa") {
+            if (++i >= argc) {
+                invalid_param = true;
+                break;
+            } else {
+                std::string value(argv[i]);
+                /**/ if (value == "distribute" || value == "" ) { params.numa = GGML_NUMA_STRATEGY_DISTRIBUTE; }
+                else if (value == "isolate")                    { params.numa = GGML_NUMA_STRATEGY_ISOLATE; }
+                else if (value == "numactl")                    { params.numa = GGML_NUMA_STRATEGY_NUMACTL; }
+                else { invalid_param = true; break; }
+            }
         } else if (arg == "-fa" || arg == "--flash-attn") {
             if (++i >= argc) {
                 invalid_param = true;
@@ -1215,6 +1229,7 @@ int main(int argc, char ** argv) {
         llama_log_set(llama_null_log_callback, NULL);
     }
     llama_backend_init();
+    llama_numa_init(params.numa);
 
     // initialize printer
     std::unique_ptr<printer> p;

examples/llava/README.md

@@ -56,7 +56,7 @@ python ./examples/llava/convert-image-encoder-to-gguf.py -m ../clip-vit-large-pa
 python ./convert.py ../llava-v1.5-7b --skip-unknown
 ```
 
-Now both the LLaMA part and the image encoder is in the `llava-v1.5-7b` directory.
+Now both the LLaMA part and the image encoder are in the `llava-v1.5-7b` directory.
 
 ## LLaVA 1.6 gguf conversion
 1) First clone a LLaVA 1.6 model:

examples/main/README.md

@@ -143,7 +143,7 @@ The `--ctx-size` option allows you to set the size of the prompt context used by
 
 ### Extended Context Size
 
-Some fine-tuned models have extended the context length by scaling RoPE. For example, if the original pre-trained model have a context length (max sequence length) of 4096 (4k) and the fine-tuned model have 32k. That is a scaling factor of 8, and should work by setting the above `--ctx-size` to 32768 (32k) and `--rope-scale` to 8.
+Some fine-tuned models have extended the context length by scaling RoPE. For example, if the original pre-trained model has a context length (max sequence length) of 4096 (4k) and the fine-tuned model has 32k. That is a scaling factor of 8, and should work by setting the above `--ctx-size` to 32768 (32k) and `--rope-scale` to 8.
 
 -   `--rope-scale N`: Where N is the linear scaling factor used by the fine-tuned model.
 
@@ -286,7 +286,7 @@ These options help improve the performance and memory usage of the LLaMA models.
 
 -   `--numa distribute`: Pin an equal proportion of the threads to the cores on each NUMA node. This will spread the load amongst all cores on the system, utilitizing all memory channels at the expense of potentially requiring memory to travel over the slow links between nodes.
 -   `--numa isolate`: Pin all threads to the NUMA node that the program starts on. This limits the number of cores and amount of memory that can be used, but guarantees all memory access remains local to the NUMA node.
--   `--numa numactl`: Pin threads to the CPUMAP that is passed to the program by starting it with the numactl utility. This is the most flexible mode, and allow arbitraty core usage patterns, for example a map that uses all the cores on one NUMA nodes, and just enough cores on a second node to saturate the inter-node memory bus.
+-   `--numa numactl`: Pin threads to the CPUMAP that is passed to the program by starting it with the numactl utility. This is the most flexible mode, and allow arbitrary core usage patterns, for example a map that uses all the cores on one NUMA nodes, and just enough cores on a second node to saturate the inter-node memory bus.
 
  These flags attempt optimizations that help on some systems with non-uniform memory access. This currently consists of one of the above strategies, and disabling prefetch and readahead for mmap. The latter causes mapped pages to be faulted in on first access instead of all at once, and in combination with pinning threads to NUMA nodes, more of the pages end up on the NUMA node where they are used. Note that if the model is already in the system page cache, for example because of a previous run without this option, this will have little effect unless you drop the page cache first. This can be done by rebooting the system or on Linux by writing '3' to '/proc/sys/vm/drop_caches' as root.
 

examples/main/main.cpp

@@ -796,7 +796,7 @@ int main(int argc, char ** argv) {
 
             // deal with end of generation tokens in interactive mode
             if (llama_token_is_eog(model, llama_sampling_last(ctx_sampling))) {
-                LOG("found EOS token\n");
+                LOG("found an EOG token\n");
 
                 if (params.interactive) {
                     if (!params.antiprompt.empty()) {

examples/server/README.md

@@ -62,6 +62,18 @@ page cache before using this. See https://github.com/ggerganov/llama.cpp/issues/
 - `--chat-template JINJA_TEMPLATE`: Set custom jinja chat template. This parameter accepts a string, not a file name.  Default: template taken from model's metadata. We only support [some pre-defined templates](https://github.com/ggerganov/llama.cpp/wiki/Templates-supported-by-llama_chat_apply_template)
 - `--log-disable`: Output logs to stdout only, not to `llama.log`. Default: enabled
 - `--log-format FORMAT`: Define the log output to FORMAT: json or text Default: `json`
+- `--rope-scaling` : RoPE scaling method. Defaults to linear unless otherwise specified by the model. Options are `none`, `linear`, `yarn`
+- `--rope-freq-base N` : RoPE frequency base (default: loaded from model)
+- `--rope-freq-scale N`: RoPE frequency scaling factor, expands context by a factor of 1/N (e.g. 0.25)
+- `--yarn-ext-factor N` : YaRN: extrapolation mix factor (Default: 1.0, 0.0 = full interpolation)
+- `--yarn-attn-factor N` : YaRN: scale sqrt(t) or attention magnitude (default: 1.0)
+- `--yarn-beta-slow N`: YaRN: High correction dim or alpha (default: 1.0)
+- `--yarn-beta-fast N`: YaRN: low correction dim or beta (default: 32.0)
+- `--pooling` : Pooling type for embeddings, use model default if unspecified. Options are `none`, `mean`, `cls`
+- `-dt N`, `--defrag-thold N`: KV cache defragmentation threshold (default: -1.0, < 0 = disabled)
+- `-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)
 
 **If compiled with `LLAMA_SERVER_SSL=ON`**
 - `--ssl-key-file FNAME`: path to file a PEM-encoded SSL private key
@@ -260,7 +272,7 @@ node index.js
 
     `logit_bias`: Modify the likelihood of a token appearing in the generated text completion. For example, use `"logit_bias": [[15043,1.0]]` to increase the likelihood of the token 'Hello', or `"logit_bias": [[15043,-1.0]]` to decrease its likelihood. Setting the value to false, `"logit_bias": [[15043,false]]` ensures that the token `Hello` is never produced. The tokens can also be represented as strings, e.g. `[["Hello, World!",-0.5]]` will reduce the likelihood of all the individual tokens that represent the string `Hello, World!`, just like the `presence_penalty` does. Default: `[]`
 
-    `n_probs`: If greater than 0, the response also contains the probabilities of top N tokens for each generated token. Default: `0`
+    `n_probs`: If greater than 0, the response also contains the probabilities of top N tokens for each generated token given the sampling settings. Note that for temperature < 0 the tokens are sampled greedily but token probabilities are still being calculated via a simple softmax of the logits without considering any other sampler settings. Default: `0`
 
     `min_keep`: If greater than 0, force samplers to return N possible tokens at minimum. Default: `0`
 

examples/server/server.cpp

@@ -2266,17 +2266,31 @@ struct server_context {
                 llama_token_data_array cur_p = { slot.ctx_sampling->cur.data(), slot.ctx_sampling->cur.size(), false };
                 result.tok = id;
 
-                const int32_t n_probs = slot.sparams.n_probs;
-                if (slot.sparams.temp <= 0 && n_probs > 0) {
-                    // for llama_sample_token_greedy we need to sort candidates
-                    llama_sample_softmax(ctx, &cur_p);
-                }
+                const size_t n_probs = std::min(cur_p.size, (size_t) slot.sparams.n_probs);
+                if (n_probs > 0) {
+                    const size_t n_considered = slot.ctx_sampling->n_considered;
 
-                for (size_t i = 0; i < std::min(cur_p.size, (size_t) n_probs); ++i) {
-                    result.probs.push_back({
-                        cur_p.data[i].id,
-                        cur_p.data[i].p
-                    });
+                    // Make sure at least n_probs top tokens are at the front of the vector:
+                    if (slot.sparams.temp == 0.0f && n_probs > n_considered) {
+                        llama_sample_top_k(ctx, &cur_p, n_probs, 0);
+                    }
+
+                    if (slot.sparams.temp == 0.0f) {
+                        // With greedy sampling the probabilities have possibly not been calculated.
+                        for (size_t i = 0; i < n_probs; ++i) {
+                            result.probs.push_back({
+                                cur_p.data[i].id,
+                                i == 0 ? 1.0f : 0.0f
+                            });
+                        }
+                    } else {
+                        for (size_t i = 0; i < n_probs; ++i) {
+                            result.probs.push_back({
+                                cur_p.data[i].id,
+                                i >= n_considered ? 0.0f : cur_p.data[i].p // Tokens filtered out due to e.g. top_k have 0 probability.
+                            });
+                        }
+                    }
                 }
 
                 if (!process_token(result, slot)) {

examples/sycl/README.md

@@ -1,6 +1,6 @@
 # llama.cpp/example/sycl
 
-This example program provide the tools for llama.cpp for SYCL on Intel GPU.
+This example program provides the tools for llama.cpp for SYCL on Intel GPU.
 
 ## Tool
 

flake.lock

@@ -5,11 +5,11 @@
         "nixpkgs-lib": "nixpkgs-lib"
       },
       "locked": {
-        "lastModified": 1712014858,
-        "narHash": "sha256-sB4SWl2lX95bExY2gMFG5HIzvva5AVMJd4Igm+GpZNw=",
+        "lastModified": 1714641030,
+        "narHash": "sha256-yzcRNDoyVP7+SCNX0wmuDju1NUCt8Dz9+lyUXEI0dbI=",
         "owner": "hercules-ci",
         "repo": "flake-parts",
-        "rev": "9126214d0a59633752a136528f5f3b9aa8565b7d",
+        "rev": "e5d10a24b66c3ea8f150e47dfdb0416ab7c3390e",
         "type": "github"
       },
       "original": {
@@ -20,11 +20,11 @@
     },
     "nixpkgs": {
       "locked": {
-        "lastModified": 1714076141,
-        "narHash": "sha256-Drmja/f5MRHZCskS6mvzFqxEaZMeciScCTFxWVLqWEY=",
+        "lastModified": 1714635257,
+        "narHash": "sha256-4cPymbty65RvF1DWQfc+Bc8B233A1BWxJnNULJKQ1EY=",
         "owner": "NixOS",
         "repo": "nixpkgs",
-        "rev": "7bb2ccd8cdc44c91edba16c48d2c8f331fb3d856",
+        "rev": "63c3a29ca82437c87573e4c6919b09a24ea61b0f",
         "type": "github"
       },
       "original": {
@@ -36,20 +36,14 @@
     },
     "nixpkgs-lib": {
       "locked": {
-        "dir": "lib",
-        "lastModified": 1711703276,
-        "narHash": "sha256-iMUFArF0WCatKK6RzfUJknjem0H9m4KgorO/p3Dopkk=",
-        "owner": "NixOS",
-        "repo": "nixpkgs",
-        "rev": "d8fe5e6c92d0d190646fb9f1056741a229980089",
-        "type": "github"
+        "lastModified": 1714640452,
+        "narHash": "sha256-QBx10+k6JWz6u7VsohfSw8g8hjdBZEf8CFzXH1/1Z94=",
+        "type": "tarball",
+        "url": "https://github.com/NixOS/nixpkgs/archive/50eb7ecf4cd0a5756d7275c8ba36790e5bd53e33.tar.gz"
       },
       "original": {
-        "dir": "lib",
-        "owner": "NixOS",
-        "ref": "nixos-unstable",
-        "repo": "nixpkgs",
-        "type": "github"
+        "type": "tarball",
+        "url": "https://github.com/NixOS/nixpkgs/archive/50eb7ecf4cd0a5756d7275c8ba36790e5bd53e33.tar.gz"
       }
     },
     "root": {

ggml-cuda.cu

@@ -113,7 +113,7 @@ static ggml_cuda_device_info ggml_cuda_init() {
     for (int id = 0; id < info.device_count; ++id) {
         int device_vmm = 0;
 
-#if !defined(GGML_USE_HIPBLAS)
+#if !defined(GGML_USE_HIPBLAS) && !defined(GGML_CUDA_NO_VMM)
         CUdevice device;
         CU_CHECK(cuDeviceGet(&device, id));
         CU_CHECK(cuDeviceGetAttribute(&device_vmm, CU_DEVICE_ATTRIBUTE_VIRTUAL_MEMORY_MANAGEMENT_SUPPORTED, device));
@@ -259,7 +259,7 @@ struct ggml_cuda_pool_leg : public ggml_cuda_pool {
 };
 
 // pool with virtual memory
-#if !defined(GGML_USE_HIPBLAS)
+#if !defined(GGML_USE_HIPBLAS) && !defined(GGML_CUDA_NO_VMM)
 struct ggml_cuda_pool_vmm : public ggml_cuda_pool {
     static const size_t CUDA_POOL_VMM_MAX_SIZE = 1ull << 35; // 32 GB
 
@@ -356,7 +356,7 @@ struct ggml_cuda_pool_vmm : public ggml_cuda_pool {
 #endif // !defined(GGML_USE_HIPBLAS)
 
 std::unique_ptr<ggml_cuda_pool> ggml_backend_cuda_context::new_pool_for_device(int device) {
-#if !defined(GGML_USE_HIPBLAS)
+#if !defined(GGML_USE_HIPBLAS) && !defined(GGML_CUDA_NO_VMM)
     if (ggml_cuda_info().devices[device].vmm) {
         return std::unique_ptr<ggml_cuda_pool>(new ggml_cuda_pool_vmm(device));
     }

grammars/README.md

@@ -51,7 +51,7 @@ single-line ::= [^\n]+ "\n"`
 
 ## Sequences and Alternatives
 
-The order of symbols in a sequence matter. For example, in `"1. " move " " move "\n"`, the `"1. "` must come before the first `move`, etc.
+The order of symbols in a sequence matters. For example, in `"1. " move " " move "\n"`, the `"1. "` must come before the first `move`, etc.
 
 Alternatives, denoted by `|`, give different sequences that are acceptable. For example, in `move ::= pawn | nonpawn | castle`, `move` can be a `pawn` move, a `nonpawn` move, or a `castle`.
 

llama.cpp

@@ -4389,6 +4389,9 @@ static void llm_load_vocab(
             } else if (
                 tokenizer_pre == "command-r") {
                 vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_COMMAND_R;
+            } else if (
+                tokenizer_pre == "olmo") {
+                vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_OLMO;
             } else {
                 throw std::runtime_error(format("unknown pre-tokenizer type: '%s'", tokenizer_pre.c_str()));
             }
@@ -12248,6 +12251,7 @@ struct llm_tokenizer_bpe {
                         });
                         break;
                     case LLAMA_VOCAB_PRE_TYPE_GPT2:
+                    case LLAMA_VOCAB_PRE_TYPE_OLMO:
                         word_collection = unicode_regex_split(text, {
                             "'s|'t|'re|'ve|'m|'ll|'d| ?\\p{L}+| ?\\p{N}+| ?[^\\s\\p{L}\\p{N}]+|\\s+(?!\\S)",
                         });

llama.h

@@ -81,6 +81,7 @@ extern "C" {
         LLAMA_VOCAB_PRE_TYPE_GPT2           = 7,
         LLAMA_VOCAB_PRE_TYPE_REFACT         = 8,
         LLAMA_VOCAB_PRE_TYPE_COMMAND_R      = 9,
+        LLAMA_VOCAB_PRE_TYPE_OLMO           = 10,
     };
 
     // note: these values should be synchronized with ggml_rope