Skip to content

initial hqq

initial hqq #9

Workflow file for this run

name: Compile main
on:
push:
branches:
- main
pull_request:
workflow_dispatch:
jobs:
run-hqq-tinystories:
strategy:
matrix:
runner: [ubuntu-latest]
runs-on: ${{matrix.runner}}
steps:
- name: Checkout repo
uses: actions/checkout@v2
- name: Setup Python
uses: actions/setup-python@v2
with:
python-version: 3.11
- name: Print machine info
run: |
uname -a
if [ $(uname -s) == Darwin ]; then
sysctl machdep.cpu.brand_string
sysctl machdep.cpu.core_count
fi
- name: Install requirements
run: |
pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cpu
pip install -r requirements.txt
pip install hqq
- name: Download checkpoints
run: |
mkdir -p checkpoints/stories15M
pushd checkpoints/stories15M
wget https://huggingface.co/karpathy/tinyllamas/resolve/main/stories15M.pt
wget https://github.com/karpathy/llama2.c/raw/master/tokenizer.model
popd
- name: Run inference
run: |
export MODEL_PATH=checkpoints/stories15M/stories15M.pt
export MODEL_NAME=stories15M
export MODEL_DIR=/tmp
echo "******************************************"
echo "******** HQQ: group-wise quantized *******"
echo "******************************************"
python generate.py --quant '{"linear:hqq" : {"group_size": 8}}' --checkpoint-path ${MODEL_PATH} --temperature 0 > ./output_eager
cat ./output_eager
python generate.py --compile --quant '{"linear:hqq" : {"group_size": 8}}' --checkpoint-path ${MODEL_PATH} --temperature 0 > ./output_compiled
cat ./output_compiled
python export.py --quant '{"embedding" : {"group_size": 8}}' --checkpoint-path ${MODEL_PATH} --output-dso-path ${MODEL_DIR}/${MODEL_NAME}.so
python generate.py --checkpoint-path ${MODEL_PATH} --temperature 0 --dso-path ${MODEL_DIR}/${MODEL_NAME}.so > ./output_aoti
cat ./output_aoti
echo "tests complete"
echo "******************************************"