Compare commits

...

3 Commits

Author SHA1 Message Date
Radoslav Gerganov f4bd8b3d26 rpc : set SO_REUSEADDR for the server socket (#7320)
ref: #7293
2024-05-17 17:25:44 +03:00
Brian 51e9d02599 Added a single test function script and fix debug-test.sh to be more robust (#7279)
* run-single-test.sh: added a single test function script and fix debug-test.sh to be more robust

* debug-test.sh: combined execute and gdb test mode via -g flag

* debug-test.sh: refactor

* debug-test: refactor for clarity

* debug-test.sh: comment style changes

* debug-test.sh: fix gdb
2024-05-17 22:40:14 +10:00
Aarni Koskela d273c1402b py : convert-hf-to-gguf-update improvements (#7340)
* convert-hf-to-gguf-update: automate updating

* convert-hf-to-gguf-update: improve download

* share requests session for performance
* create directories only when needed, don't skip downloads when empty directory encountered
* be more graceful about errors
2024-05-17 15:11:45 +03:00
6 changed files with 252 additions and 144 deletions
+39 -48
View File
@@ -20,11 +20,13 @@
# - Update llama.cpp with the new pre-tokenizer if necessary
#
# TODO: generate tokenizer tests for llama.cpp
# TODO: automate the update of convert-hf-to-gguf.py
#
import logging
import os
import pathlib
import re
import requests
import sys
import json
@@ -35,6 +37,7 @@ from transformers import AutoTokenizer
logging.basicConfig(level=logging.DEBUG)
logger = logging.getLogger("convert-hf-to-gguf-update")
sess = requests.Session()
class TOKENIZER_TYPE(IntEnum):
@@ -79,63 +82,44 @@ models = [
{"name": "jina-v2-de", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/jinaai/jina-embeddings-v2-base-de", },
]
# make directory "models/tokenizers" if it doesn't exist
if not os.path.exists("models/tokenizers"):
os.makedirs("models/tokenizers")
def download_file_with_auth(url, token, save_path):
headers = {"Authorization": f"Bearer {token}"}
response = requests.get(url, headers=headers)
if response.status_code == 200:
with open(save_path, 'wb') as f:
f.write(response.content)
logger.info(f"File {save_path} downloaded successfully")
else:
logger.info(f"Failed to download file. Status code: {response.status_code}")
response = sess.get(url, headers=headers)
response.raise_for_status()
os.makedirs(os.path.dirname(save_path), exist_ok=True)
with open(save_path, 'wb') as f:
f.write(response.content)
logger.info(f"File {save_path} downloaded successfully")
# download the tokenizer models
for model in models:
def download_model(model):
name = model["name"]
repo = model["repo"]
tokt = model["tokt"]
if not os.path.exists(f"models/tokenizers/{name}"):
os.makedirs(f"models/tokenizers/{name}")
else:
logger.info(f"Directory models/tokenizers/{name} already exists - skipping")
continue
logger.info(f"Downloading {name} to models/tokenizers/{name}")
url = f"{repo}/raw/main/config.json"
save_path = f"models/tokenizers/{name}/config.json"
download_file_with_auth(url, token, save_path)
url = f"{repo}/raw/main/tokenizer.json"
save_path = f"models/tokenizers/{name}/tokenizer.json"
download_file_with_auth(url, token, save_path)
# if downloaded file is less than 1KB, we likely need to download an LFS instead
if os.path.getsize(save_path) < 1024:
# remove the file
os.remove(save_path)
url = f"{repo}/resolve/main/tokenizer.json"
save_path = f"models/tokenizers/{name}/tokenizer.json"
download_file_with_auth(url, token, save_path)
os.makedirs(f"models/tokenizers/{name}", exist_ok=True)
files = ["config.json", "tokenizer.json", "tokenizer_config.json"]
if tokt == TOKENIZER_TYPE.SPM:
url = f"{repo}/resolve/main/tokenizer.model"
save_path = f"models/tokenizers/{name}/tokenizer.model"
download_file_with_auth(url, token, save_path)
files.append("tokenizer.model")
for file in files:
save_path = f"models/tokenizers/{name}/{file}"
if os.path.isfile(save_path):
logger.info(f"{name}: File {save_path} already exists - skipping")
continue
download_file_with_auth(f"{repo}/resolve/main/{file}", token, save_path)
for model in models:
try:
download_model(model)
except Exception as e:
logger.error(f"Failed to download model {model['name']}. Error: {e}")
url = f"{repo}/raw/main/tokenizer_config.json"
save_path = f"models/tokenizers/{name}/tokenizer_config.json"
download_file_with_auth(url, token, save_path)
# generate the source code for the convert-hf-to-gguf.py:get_vocab_base_pre() function:
# TODO: auto-update convert-hf-to-gguf.py with the generated function
src_ifs = ""
for model in models:
@@ -224,11 +208,18 @@ src_func = f"""
return res
"""
print(src_func) # noqa: NP100
convert_py_pth = pathlib.Path("convert-hf-to-gguf.py")
convert_py = convert_py_pth.read_text()
convert_py = re.sub(
r"(# Marker: Start get_vocab_base_pre)(.+?)( +# Marker: End get_vocab_base_pre)",
lambda m: m.group(1) + src_func + m.group(3),
convert_py,
flags=re.DOTALL | re.MULTILINE,
)
logger.info("\n")
logger.info("!!! Copy-paste the function above into convert-hf-to-gguf.py !!!")
logger.info("\n")
convert_py_pth.write_text(convert_py)
logger.info("+++ convert-hf-to-gguf.py was updated")
# generate tests for each tokenizer model
+2
View File
@@ -402,6 +402,7 @@ class Model:
# 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
def get_vocab_base_pre(self, tokenizer) -> str:
# encoding this string and hashing the resulting tokens would (hopefully) give us a unique identifier that
# is specific for the BPE pre-tokenizer used by the model
@@ -489,6 +490,7 @@ class Model:
logger.debug(f"chkhsh: {chkhsh}")
return res
# Marker: End get_vocab_base_pre
def _set_vocab_gpt2(self) -> None:
tokens, toktypes, tokpre = self.get_vocab_base()
+20 -4
View File
@@ -1,6 +1,6 @@
# Debugging Tests Tips
## How to run & debug a specific test without anything else to keep the feedback loop short?
## How to run & execute or debug a specific test without anything else to keep the feedback loop short?
There is a script called debug-test.sh in the scripts folder whose parameter takes a REGEX and an optional test number.
@@ -10,13 +10,27 @@ For example, running the following command will output an interactive list from
It will then build & run in the debugger for you.
To just execute a test and get back a PASS or FAIL message run:
```bash
./scripts/debug-test.sh test-tokenizer
```
To test in GDB use the `-g` flag to enable gdb test mode.
```bash
./scripts/debug-test.sh -g test-tokenizer
# Once in the debugger, i.e. at the chevrons prompt, setting a breakpoint could be as follows:
>>> b main
```
To speed up the testing loop, if you know your test number you can just run it similar to below:
```bash
./scripts/debug-test.sh test 23
```
For further reference use `debug-test.sh -h` to print help.
&nbsp;
@@ -41,7 +55,7 @@ cmake -DCMAKE_BUILD_TYPE=Debug -DLLAMA_CUDA=1 -DLLAMA_FATAL_WARNINGS=ON ..
make -j
```
#### Step 3.1: Identify Test Command for Debugging
#### Step 3: Find all tests available that matches REGEX
The output of this command will give you the command & arguments needed to run GDB.
@@ -69,11 +83,13 @@ Labels: main
...
```
So for test #1 we can tell these two pieces of relevant information:
#### Step 4: Identify Test Command for Debugging
So for test #1 above we can tell these two pieces of relevant information:
* Test Binary: `~/llama.cpp/build-ci-debug/bin/test-tokenizer-0`
* Test GGUF Model: `~/llama.cpp/tests/../models/ggml-vocab-llama-spm.gguf`
#### Step 3.2: Run GDB on test command
#### Step 5: Run GDB on test command
Based on the ctest 'test command' report above we can then run a gdb session via this command below:
+4
View File
@@ -56,6 +56,10 @@ static bool rpc_server_params_parse(int argc, char ** argv, rpc_server_params &
} else if (arg == "-h" || arg == "--help") {
print_usage(argc, argv, params);
exit(0);
} else {
fprintf(stderr, "error: unknown argument: %s\n", arg.c_str());
print_usage(argc, argv, params);
exit(0);
}
}
return true;
+11 -2
View File
@@ -134,7 +134,13 @@ static bool set_no_delay(sockfd_t sockfd) {
int flag = 1;
// set TCP_NODELAY to disable Nagle's algorithm
int ret = setsockopt(sockfd, IPPROTO_TCP, TCP_NODELAY, (char *)&flag, sizeof(int));
return ret >= 0;
return ret == 0;
}
static bool set_reuse_addr(sockfd_t sockfd) {
int flag = 1;
int ret = setsockopt(sockfd, SOL_SOCKET, SO_REUSEADDR, (char *)&flag, sizeof(int));
return ret == 0;
}
static std::shared_ptr<socket_t> socket_connect(const char * host, int port) {
@@ -181,7 +187,10 @@ static std::shared_ptr<socket_t> create_server_socket(const char * host, int por
if (sock == nullptr) {
return nullptr;
}
if (!set_reuse_addr(sockfd)) {
fprintf(stderr, "Failed to set SO_REUSEADDR\n");
return nullptr;
}
struct sockaddr_in serv_addr;
serv_addr.sin_family = AF_INET;
serv_addr.sin_addr.s_addr = inet_addr(host);
+176 -90
View File
@@ -1,117 +1,203 @@
#!/bin/bash
test_suite=${1:-}
test_number=${2:-}
PROG=${0##*/}
build_dir="build-ci-debug"
if [ x"$1" = x"-h" ] || [ x"$1" = x"--help" ]; then
echo "Usage: $PROG [OPTION]... <test_regex> (test_number)"
echo "Debug specific ctest program."
echo
echo "Options:"
echo " -h, --help Display this help and exit"
echo
echo "Arguments:"
echo " <test_regex> (Mandatory) Supply one regex to the script to filter tests"
echo " (test_number) (Optional) Test number to run a specific test"
echo
echo "Example:"
echo " $PROG test-tokenizer"
echo " $PROG test-tokenizer 3"
echo
exit 0
fi
# Print Color Commands
red=$(tput setaf 1)
green=$(tput setaf 2)
yellow=$(tput setaf 3)
blue=$(tput setaf 4)
magenta=$(tput setaf 5)
cyan=$(tput setaf 6)
normal=$(tput sgr0)
# Function to select and debug a test
function select_test() {
test_suite=${1:-test}
test_number=${2:-}
# Sanity Check If Tests Is Detected
printf "\n\nGathering tests that fit REGEX: ${test_suite} ...\n"
tests=($(ctest -R ${test_suite} -V -N | grep -E " +Test +#[0-9]+*" | cut -d':' -f2 | awk '{$1=$1};1'))
if [ ${#tests[@]} -eq 0 ]
then
echo "No tests avaliable... check your compliation process..."
echo "Exiting."
exit 1
fi
# Print Help Message
####################
if [ -z $test_number ]
then
# List out avaliable tests
printf "Which test would you like to debug?\n"
id=0
for s in "${tests[@]}"
do
echo "Test# ${id}"
echo " $s"
((id++))
done
print_full_help() {
cat << EOF
Usage: $PROG [OPTION]... <test_regex> (test_number)
Debug specific ctest program.
# Prompt user which test they wanted to run
printf "\nRun test#? "
read test_number
else
printf "\nUser Already Requested #${test_number}"
fi
Options:
-h, --help display this help and exit
-g run in gdb mode
# Start GDB with the requested test binary and arguments
printf "Debugging(GDB) test: ${tests[test_number]}\n"
# Change IFS (Internal Field Separator)
sIFS=$IFS
IFS=$'\n'
Arguments:
<test_regex> (Mandatory) Supply one regex to the script to filter tests
(test_number) (Optional) Test number to run a specific test
# Get test args
gdb_args=($(ctest -R ${test_suite} -V -N | grep "Test command" | cut -d':' -f3 | awk '{$1=$1};1' ))
IFS=$sIFS
printf "Debug arguments: ${gdb_args[test_number]}\n\n"
# Expand paths if needed
args=()
for x in $(echo ${gdb_args[test_number]} | sed -e 's/"\/\<//' -e 's/\>"//')
do
args+=($(echo $x | sed -e 's/.*\/..\//..\//'))
done
# Execute debugger
echo "gdb args: ${args[@]}"
gdb --args ${args[@]}
Example:
$PROG test-tokenizer
$PROG test-tokenizer 3
EOF
}
abort() {
echo "Error: $1" >&2
cat << EOF >&2
Usage: $PROG [OPTION]... <test_regex> (test_number)
Debug specific ctest program.
Refer to --help for full instructions.
EOF
exit 1
}
# Dependency Sanity Check
#########################
check_dependency() {
command -v "$1" >/dev/null 2>&1 || {
abort "$1 is required but not found. Please install it and try again."
}
}
check_dependency ctest
check_dependency cmake
# Step 0: Check the args
if [ -z "$test_suite" ]
then
echo "Usage: $PROG [OPTION]... <test_regex> (test_number)"
echo "Supply one regex to the script to filter tests,"
echo "and optionally a test number to run a specific test."
echo "Use --help flag for full instructions"
exit 1
########################
if [ x"$1" = x"-h" ] || [ x"$1" = x"--help" ]; then
print_full_help >&2
exit 0
fi
# Parse command-line options
gdb_mode=false
while getopts "g" opt; do
case $opt in
g)
gdb_mode=true
echo "gdb_mode Mode Enabled"
;;
esac
done
# Shift the option parameters
shift $((OPTIND - 1))
# Positionial Argument Processing : <test_regex>
if [ -z "${1}" ]; then
abort "Test regex is required"
else
test_suite=${1:-}
fi
# Positionial Argument Processing : (test_number)
test_number=${2:-}
# Step 1: Reset and Setup folder context
########################################
## Sanity check that we are actually in a git repo
repo_root=$(git rev-parse --show-toplevel)
if [ ! -d "$repo_root" ]; then
echo "Error: Not in a Git repository."
exit 1
abort "Not in a Git repository."
fi
## Reset folder to root context of git repo
pushd "$repo_root" || exit 1
## Reset folder to root context of git repo and Create and enter build directory
pushd "$repo_root"
rm -rf "$build_dir" && mkdir "$build_dir" || abort "Failed to make $build_dir"
## Create and enter build directory
rm -rf "$build_dir" && mkdir "$build_dir" || exit 1
# Step 2: Setup Build Environment and Compile Test Binaries
cmake -B "./$build_dir" -DCMAKE_BUILD_TYPE=Debug -DLLAMA_CUDA=1 -DLLAMA_FATAL_WARNINGS=ON || exit 1
pushd "$build_dir" && make -j || exit 1
###########################################################
# Step 3: Debug the Test
select_test "$test_suite" "$test_number"
# Note: test-eval-callback requires -DLLAMA_CURL
cmake -B "./$build_dir" -DCMAKE_BUILD_TYPE=Debug -DLLAMA_CUDA=1 -DLLAMA_CURL=1 || abort "Failed to build enviroment"
pushd "$build_dir"
make -j || abort "Failed to compile"
popd > /dev/null || exit 1
# Step 4: Return to the directory from which the user ran the command.
popd || exit 1
popd || exit 1
popd || exit 1
# Step 3: Find all tests available that matches REGEX
####################################################
# Ctest Gather Tests
# `-R test-tokenizer` : looks for all the test files named `test-tokenizer*` (R=Regex)
# `-N` : "show-only" disables test execution & shows test commands that you can feed to GDB.
# `-V` : Verbose Mode
printf "\n\nGathering tests that fit REGEX: ${test_suite} ...\n"
pushd "$build_dir"
tests=($(ctest -R ${test_suite} -V -N | grep -E " +Test +#[0-9]+*" | cut -d':' -f2 | awk '{$1=$1};1'))
if [ ${#tests[@]} -eq 0 ]; then
abort "No tests avaliable... check your compliation process..."
fi
popd > /dev/null || exit 1
# Step 4: Identify Test Command for Debugging
#############################################
# Select test number
if [ -z $test_number ]; then
# List out avaliable tests
printf "Which test would you like to debug?\n"
id=0
for s in "${tests[@]}"
do
echo "Test# ${id}"
echo " $s"
((id++))
done
# Prompt user which test they wanted to run
printf "\nRun test#? "
read test_number
else
printf "\nUser Already Requested #${test_number}\n"
fi
# Grab all tests commands
pushd "$build_dir"
sIFS=$IFS # Save Initial IFS (Internal Field Separator)
IFS=$'\n' # Change IFS (Internal Field Separator) (So we split ctest output by newline rather than by spaces)
test_args=($(ctest -R ${test_suite} -V -N | grep "Test command" | cut -d':' -f3 | awk '{$1=$1};1' )) # Get test args
IFS=$sIFS # Reset IFS (Internal Field Separator)
popd > /dev/null || exit 1
# Grab specific test command
single_test_name="${tests[test_number]}"
single_test_command="${test_args[test_number]}"
# Step 5: Execute or GDB Debug
##############################
printf "${magenta}Running Test #${test_number}: ${single_test_name}${normal}\n"
printf "${cyan}single_test_command: ${single_test_command}${normal}\n"
if [ "$gdb_mode" = "true" ]; then
# Execute debugger
pushd "$repo_root" || exit 1
eval "gdb --args ${single_test_command}"
popd > /dev/null || exit 1
else
# Execute Test
pushd "$repo_root" || exit 1
eval "${single_test_command}"
exit_code=$?
popd > /dev/null || exit 1
# Print Result
printf "${blue}Ran Test #${test_number}: ${single_test_name}${normal}\n"
printf "${yellow}Command: ${single_test_command}${normal}\n"
if [ $exit_code -eq 0 ]; then
printf "${green}TEST PASS${normal}\n"
else
printf "${red}TEST FAIL${normal}\n"
fi
fi
# Return to the directory from which the user ran the command.
popd > /dev/null || exit 1