mirror of
https://github.com/ggerganov/llama.cpp.git
synced 2024-12-25 02:44:36 +00:00
Fix --split-max-size (#6655)
* Fix --split-max-size Byte size calculation was done on int and overflowed. * add tests.sh * add examples test scripts to ci run Will autodiscover examples/*/tests.sh scripts and run them. * move WORK_PATH to a subdirectory * clean up before and after test * explicitly define which scripts to run * add --split-max-size to readme
This commit is contained in:
parent
e689fc4e91
commit
8800226d65
49
ci/run.sh
49
ci/run.sh
@ -153,6 +153,52 @@ function gg_sum_ctest_release {
|
|||||||
gg_printf '```\n'
|
gg_printf '```\n'
|
||||||
}
|
}
|
||||||
|
|
||||||
|
# test_scripts_debug
|
||||||
|
|
||||||
|
function gg_run_test_scripts_debug {
|
||||||
|
cd ${SRC}
|
||||||
|
|
||||||
|
set -e
|
||||||
|
|
||||||
|
(cd ./examples/gguf-split && time bash tests.sh "$SRC/build-ci-debug/bin" "$MNT/models") 2>&1 | tee -a $OUT/${ci}-scripts.log
|
||||||
|
|
||||||
|
set +e
|
||||||
|
}
|
||||||
|
|
||||||
|
function gg_sum_test_scripts_debug {
|
||||||
|
gg_printf '### %s\n\n' "${ci}"
|
||||||
|
|
||||||
|
gg_printf 'Runs test scripts in debug mode\n'
|
||||||
|
gg_printf '- status: %s\n' "$(cat $OUT/${ci}.exit)"
|
||||||
|
gg_printf '```\n'
|
||||||
|
gg_printf '%s\n' "$(cat $OUT/${ci}-scripts.log)"
|
||||||
|
gg_printf '```\n'
|
||||||
|
gg_printf '\n'
|
||||||
|
}
|
||||||
|
|
||||||
|
# test_scripts_release
|
||||||
|
|
||||||
|
function gg_run_test_scripts_release {
|
||||||
|
cd ${SRC}
|
||||||
|
|
||||||
|
set -e
|
||||||
|
|
||||||
|
(cd ./examples/gguf-split && time bash tests.sh "$SRC/build-ci-release/bin" "$MNT/models") 2>&1 | tee -a $OUT/${ci}-scripts.log
|
||||||
|
|
||||||
|
set +e
|
||||||
|
}
|
||||||
|
|
||||||
|
function gg_sum_test_scripts_release {
|
||||||
|
gg_printf '### %s\n\n' "${ci}"
|
||||||
|
|
||||||
|
gg_printf 'Runs test scripts in release mode\n'
|
||||||
|
gg_printf '- status: %s\n' "$(cat $OUT/${ci}.exit)"
|
||||||
|
gg_printf '```\n'
|
||||||
|
gg_printf '%s\n' "$(cat $OUT/${ci}-scripts.log)"
|
||||||
|
gg_printf '```\n'
|
||||||
|
gg_printf '\n'
|
||||||
|
}
|
||||||
|
|
||||||
function gg_get_model {
|
function gg_get_model {
|
||||||
local gguf_3b="$MNT/models/open-llama/3B-v2/ggml-model-f16.gguf"
|
local gguf_3b="$MNT/models/open-llama/3B-v2/ggml-model-f16.gguf"
|
||||||
local gguf_7b="$MNT/models/open-llama/7B-v2/ggml-model-f16.gguf"
|
local gguf_7b="$MNT/models/open-llama/7B-v2/ggml-model-f16.gguf"
|
||||||
@ -642,6 +688,9 @@ test $ret -eq 0 && gg_run ctest_release
|
|||||||
if [ -z ${GG_BUILD_LOW_PERF} ]; then
|
if [ -z ${GG_BUILD_LOW_PERF} ]; then
|
||||||
test $ret -eq 0 && gg_run embd_bge_small
|
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_VRAM_GB} ] || [ ${GG_BUILD_VRAM_GB} -ge 8 ]; then
|
if [ -z ${GG_BUILD_VRAM_GB} ] || [ ${GG_BUILD_VRAM_GB} -ge 8 ]; then
|
||||||
if [ -z ${GG_BUILD_CUDA} ]; then
|
if [ -z ${GG_BUILD_CUDA} ]; then
|
||||||
test $ret -eq 0 && gg_run open_llama_3b_v2
|
test $ret -eq 0 && gg_run open_llama_3b_v2
|
||||||
|
@ -5,5 +5,6 @@ CLI to split / merge GGUF files.
|
|||||||
**Command line options:**
|
**Command line options:**
|
||||||
|
|
||||||
- `--split`: split GGUF to multiple GGUF, default operation.
|
- `--split`: split GGUF to multiple GGUF, default operation.
|
||||||
|
- `--split-max-size`: max size per split in `M` or `G`, f.ex. `500M` or `2G`.
|
||||||
- `--split-max-tensors`: maximum tensors in each split: default(128)
|
- `--split-max-tensors`: maximum tensors in each split: default(128)
|
||||||
- `--merge`: merge multiple GGUF to a single GGUF.
|
- `--merge`: merge multiple GGUF to a single GGUF.
|
||||||
|
@ -59,10 +59,10 @@ static size_t split_str_to_n_bytes(std::string str) {
|
|||||||
int n;
|
int n;
|
||||||
if (str.back() == 'M') {
|
if (str.back() == 'M') {
|
||||||
sscanf(str.c_str(), "%d", &n);
|
sscanf(str.c_str(), "%d", &n);
|
||||||
n_bytes = n * 1024 * 1024; // megabytes
|
n_bytes = (size_t)n * 1024 * 1024; // megabytes
|
||||||
} else if (str.back() == 'G') {
|
} else if (str.back() == 'G') {
|
||||||
sscanf(str.c_str(), "%d", &n);
|
sscanf(str.c_str(), "%d", &n);
|
||||||
n_bytes = n * 1024 * 1024 * 1024; // gigabytes
|
n_bytes = (size_t)n * 1024 * 1024 * 1024; // gigabytes
|
||||||
} else {
|
} else {
|
||||||
throw std::invalid_argument("error: supported units are M (megabytes) or G (gigabytes), but got: " + std::string(1, str.back()));
|
throw std::invalid_argument("error: supported units are M (megabytes) or G (gigabytes), but got: " + std::string(1, str.back()));
|
||||||
}
|
}
|
||||||
|
89
examples/gguf-split/tests.sh
Normal file
89
examples/gguf-split/tests.sh
Normal file
@ -0,0 +1,89 @@
|
|||||||
|
#!/bin/bash
|
||||||
|
|
||||||
|
set -eu
|
||||||
|
|
||||||
|
if [ $# -lt 1 ]
|
||||||
|
then
|
||||||
|
echo "usage: $0 path_to_build_binary [path_to_temp_folder]"
|
||||||
|
echo "example: $0 ../../build/bin ../../tmp"
|
||||||
|
exit 1
|
||||||
|
fi
|
||||||
|
|
||||||
|
if [ $# -gt 1 ]
|
||||||
|
then
|
||||||
|
TMP_DIR=$2
|
||||||
|
else
|
||||||
|
TMP_DIR=/tmp
|
||||||
|
fi
|
||||||
|
|
||||||
|
set -x
|
||||||
|
|
||||||
|
SPLIT=$1/gguf-split
|
||||||
|
MAIN=$1/main
|
||||||
|
WORK_PATH=$TMP_DIR/gguf-split
|
||||||
|
CUR_DIR=$(pwd)
|
||||||
|
|
||||||
|
mkdir -p "$WORK_PATH"
|
||||||
|
|
||||||
|
# Clean up in case of previously failed test
|
||||||
|
rm -f $WORK_PATH/ggml-model-split*.gguf $WORK_PATH/ggml-model-merge*.gguf
|
||||||
|
|
||||||
|
# 1. Get a model
|
||||||
|
(
|
||||||
|
cd $WORK_PATH
|
||||||
|
"$CUR_DIR"/../../scripts/hf.sh --repo ggml-org/gemma-1.1-2b-it-Q8_0-GGUF --file gemma-1.1-2b-it.Q8_0.gguf
|
||||||
|
)
|
||||||
|
echo PASS
|
||||||
|
|
||||||
|
# 2. Split with max tensors strategy
|
||||||
|
$SPLIT --split-max-tensors 28 $WORK_PATH/gemma-1.1-2b-it.Q8_0.gguf $WORK_PATH/ggml-model-split
|
||||||
|
echo PASS
|
||||||
|
echo
|
||||||
|
|
||||||
|
# 2b. Test the sharded model is loading properly
|
||||||
|
$MAIN --model $WORK_PATH/ggml-model-split-00001-of-00006.gguf --random-prompt --n-predict 32
|
||||||
|
echo PASS
|
||||||
|
echo
|
||||||
|
|
||||||
|
# 3. Merge
|
||||||
|
$SPLIT --merge $WORK_PATH/ggml-model-split-00001-of-00006.gguf $WORK_PATH/ggml-model-merge.gguf
|
||||||
|
echo PASS
|
||||||
|
echo
|
||||||
|
|
||||||
|
# 3b. Test the merged model is loading properly
|
||||||
|
$MAIN --model $WORK_PATH/ggml-model-merge.gguf --random-prompt --n-predict 32
|
||||||
|
echo PASS
|
||||||
|
echo
|
||||||
|
|
||||||
|
# 4. Split with no tensor in metadata
|
||||||
|
#$SPLIT --split-max-tensors 32 --no-tensor-in-metadata $WORK_PATH/ggml-model-merge.gguf $WORK_PATH/ggml-model-split-32-tensors
|
||||||
|
#echo PASS
|
||||||
|
#echo
|
||||||
|
|
||||||
|
# 4b. Test the sharded model is loading properly
|
||||||
|
#$MAIN --model $WORK_PATH/ggml-model-split-32-tensors-00001-of-00006.gguf --random-prompt --n-predict 32
|
||||||
|
#echo PASS
|
||||||
|
#echo
|
||||||
|
|
||||||
|
# 5. Merge
|
||||||
|
#$SPLIT --merge $WORK_PATH/ggml-model-split-32-tensors-00001-of-00006.gguf $WORK_PATH/ggml-model-merge-2.gguf
|
||||||
|
#echo PASS
|
||||||
|
#echo
|
||||||
|
|
||||||
|
# 5b. Test the merged model is loading properly
|
||||||
|
#$MAIN --model $WORK_PATH/ggml-model-merge-2.gguf --random-prompt --n-predict 32
|
||||||
|
#echo PASS
|
||||||
|
#echo
|
||||||
|
|
||||||
|
# 6. Split with size strategy
|
||||||
|
$SPLIT --split-max-size 2G $WORK_PATH/ggml-model-merge.gguf $WORK_PATH/ggml-model-split-2G
|
||||||
|
echo PASS
|
||||||
|
echo
|
||||||
|
|
||||||
|
# 6b. Test the sharded model is loading properly
|
||||||
|
$MAIN --model $WORK_PATH/ggml-model-split-2G-00001-of-00002.gguf --random-prompt --n-predict 32
|
||||||
|
echo PASS
|
||||||
|
echo
|
||||||
|
|
||||||
|
# Clean up
|
||||||
|
rm -f $WORK_PATH/ggml-model-split*.gguf $WORK_PATH/ggml-model-merge*.gguf
|
Loading…
Reference in New Issue
Block a user