llama.cpp/scripts/check-requirements.sh
Georgi Gerganov f4ab2a4147
llama : fix BPE pre-tokenization (#6920)
* merged the changes from deepseeker models to main branch

* Moved regex patterns to unicode.cpp and updated unicode.h

* Moved header files

* Resolved issues

* added and refactored unicode_regex_split and related functions

* Updated/merged the deepseek coder pr

* Refactored code

* Adding unicode regex mappings

* Adding unicode regex function

* Added needed functionality, testing remains

* Fixed issues

* Fixed issue with gpt2 regex custom preprocessor

* unicode : fix? unicode_wstring_to_utf8

* lint : fix whitespaces

* tests : add tokenizer tests for numbers

* unicode : remove redundant headers

* tests : remove and rename tokenizer test scripts

* tests : add sample usage

* gguf-py : reader prints warnings on duplicate keys

* llama : towards llama3 tokenization support (wip)

* unicode : shot in the dark to fix tests on Windows

* unicode : first try custom implementations

* convert : add "tokenizer.ggml.pre" GGUF KV (wip)

* llama : use new pre-tokenizer type

* convert : fix pre-tokenizer type writing

* lint : fix

* make : add test-tokenizer-0-llama-v3

* wip

* models : add llama v3 vocab file

* llama : adapt punctuation regex + add llama 3 regex

* minor

* unicode : set bomb

* unicode : set bomb

* unicode : always use std::wregex

* unicode : support \p{N}, \p{L} and \p{P} natively

* unicode : try fix windows

* unicode : category support via std::regex

* unicode : clean-up

* unicode : simplify

* convert : add convert-hf-to-gguf-update.py

ggml-ci

* lint : update

* convert : add falcon

ggml-ci

* unicode : normalize signatures

* lint : fix

* lint : fix

* convert : remove unused functions

* convert : add comments

* convert : exercise contractions

ggml-ci

* lint : fix

* cmake : refactor test targets

* tests : refactor vocab tests

ggml-ci

* tests : add more vocabs and tests

ggml-ci

* unicode : cleanup

* scripts : ignore new update script in check-requirements.sh

* models : add phi-3, mpt, gpt-2, starcoder

* tests : disable obsolete

ggml-ci

* tests : use faster bpe test

ggml-ci

* llama : more prominent warning for old BPE models

* tests : disable test-tokenizer-1-bpe due to slowness

ggml-ci

---------

Co-authored-by: Jaggzh <jaggz.h@gmail.com>
Co-authored-by: Kazim Abrar Mahi <kazimabrarmahi135@gmail.com>
2024-04-29 16:58:41 +03:00

180 lines
4.3 KiB
Bash
Executable File

#!/bin/bash
set -euo pipefail
#
# check-requirements.sh checks all requirements files for each top-level
# convert*.py script.
#
# WARNING: This is quite IO intensive, because a fresh venv is set up for every
# python script. As of 2023-12-22, this writes ~2.7GB of data. An adequately
# sized tmpfs /tmp or ramdisk is recommended if running this frequently.
#
# usage: check-requirements.sh [<working_dir>]
# check-requirements.sh nocleanup [<working_dir>]
#
# where:
# - <working_dir> is a directory that can be used as the base for
# setting up the venvs. Defaults to `/tmp`.
# - 'nocleanup' as the first argument will disable automatic cleanup
# of the files created by this script.
#
# requires:
# - bash >= 3.2.57
# - shellcheck
#
# For each script, it creates a fresh venv, `pip install`s the requirements, and
# finally imports the python script to check for `ImportError`.
#
log() {
local level=$1 msg=$2
printf >&2 '%s: %s\n' "$level" "$msg"
}
debug() {
log DEBUG "$@"
}
info() {
log INFO "$@"
}
fatal() {
log FATAL "$@"
exit 1
}
cleanup() {
if [[ -n ${workdir+x} && -d $workdir && -w $workdir ]]; then
info "Removing $workdir"
local count=0
rm -rfv -- "$workdir" | while read -r; do
if (( count++ > 750 )); then
printf .
count=0
fi
done
printf '\n'
info "Removed $workdir"
fi
}
do_cleanup=1
if [[ ${1-} == nocleanup ]]; then
do_cleanup=0; shift
fi
if (( do_cleanup )); then
trap exit INT TERM
trap cleanup EXIT
fi
this=$(realpath -- "$0"); readonly this
cd "$(dirname "$this")/.." # PWD should stay in llama.cpp project directory
shellcheck "$this"
readonly reqs_dir=requirements
if [[ ${1+x} ]]; then
tmp_dir=$(realpath -- "$1")
if [[ ! ( -d $tmp_dir && -w $tmp_dir ) ]]; then
fatal "$tmp_dir is not a writable directory"
fi
else
tmp_dir=/tmp
fi
workdir=$(mktemp -d "$tmp_dir/check-requirements.XXXX"); readonly workdir
info "Working directory: $workdir"
check_requirements() {
local reqs=$1
info "$reqs: beginning check"
pip --disable-pip-version-check install -qr "$reqs"
info "$reqs: OK"
}
check_convert_script() {
local py=$1 # e.g. ./convert-hf-to-gguf.py
local pyname=${py##*/} # e.g. convert-hf-to-gguf.py
pyname=${pyname%.py} # e.g. convert-hf-to-gguf
info "$py: beginning check"
local reqs="$reqs_dir/requirements-$pyname.txt"
if [[ ! -r $reqs ]]; then
fatal "$py missing requirements. Expected: $reqs"
fi
local venv="$workdir/$pyname-venv"
python3 -m venv "$venv"
(
# shellcheck source=/dev/null
source "$venv/bin/activate"
check_requirements "$reqs"
python - "$py" "$pyname" <<'EOF'
import sys
from importlib.machinery import SourceFileLoader
py, pyname = sys.argv[1:]
SourceFileLoader(pyname, py).load_module()
EOF
)
if (( do_cleanup )); then
rm -rf -- "$venv"
fi
info "$py: imports OK"
}
readonly ignore_eq_eq='check_requirements: ignore "=="'
for req in "$reqs_dir"/*; do
# Check that all sub-requirements are added to top-level requirements.txt
if ! grep -qF "$req" requirements.txt; then
fatal "$req needs to be added to requirements.txt"
fi
# Make sure exact release versions aren't being pinned in the requirements
# Filters out the ignore string
if grep -vF "$ignore_eq_eq" "$req" | grep -q '=='; then
tab=$'\t'
cat >&2 <<EOF
FATAL: Avoid pinning exact package versions. Use '~=' instead.
You can suppress this error by appending the following to the line:
$tab# $ignore_eq_eq
EOF
exit 1
fi
done
all_venv="$workdir/all-venv"
python3 -m venv "$all_venv"
(
# shellcheck source=/dev/null
source "$all_venv/bin/activate"
check_requirements requirements.txt
)
if (( do_cleanup )); then
rm -rf -- "$all_venv"
fi
check_convert_script convert.py
for py in convert-*.py; do
# skip convert-hf-to-gguf-update.py
# TODO: the check is failing for some reason:
# https://github.com/ggerganov/llama.cpp/actions/runs/8875330981/job/24364557177?pr=6920
[[ $py == convert-hf-to-gguf-update.py ]] && continue
check_convert_script "$py"
done
info 'Done! No issues found.'