Docs #130
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| name: Compile main using GGUF | |
| on: | |
| push: | |
| branches: | |
| - main | |
| pull_request: | |
| workflow_dispatch: | |
| jobs: | |
| run-tinystories: | |
| strategy: | |
| matrix: | |
| runner: [macos-14] | |
| 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 gguf | |
| pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cpu | |
| pip install -r requirements.txt | |
| - name: Download GGUF | |
| run: | | |
| mkdir gguf_files | |
| export GGUF_PATH=gguf_files/TinyLlama-1.1B-openorca.Q4_0.gguf | |
| export TOKENIZER_PATH=gguf_files/tokenizer.model | |
| wget -O ${GGUF_PATH} "https://huggingface.co/TheBloke/TinyLlama-1.1B-1T-OpenOrca-GGUF/resolve/main/tinyllama-1.1b-1t-openorca.Q4_0.gguf?download=true" | |
| wget -O ${TOKENIZER_PATH} https://github.com/karpathy/llama2.c/raw/master/tokenizer.model | |
| - name: Run inference | |
| run: | | |
| export GGUF_PATH=gguf_files/TinyLlama-1.1B-openorca.Q4_0.gguf | |
| export TOKENIZER_PATH=gguf_files/tokenizer.model | |
| export MODEL_NAME=TinyLlama-1.1B-openorca.Q4_0.gguf | |
| export MODEL_DIR=/tmp | |
| echo "******************************************" | |
| echo "******* Embed: not quantized *************" | |
| echo "******************************************" | |
| echo "Running eager" | |
| python generate.py --gguf-path ${GGUF_PATH} --tokenizer-path ${TOKENIZER_PATH} --max-new-tokens 20 --temperature 0 > ./output_eager | |
| cat ./output_eager | |
| echo "Running compiled" | |
| python generate.py --compile --gguf-path ${GGUF_PATH} --tokenizer-path ${TOKENIZER_PATH} --max-new-tokens 20 --temperature 0 > ./output_compiled | |
| cat ./output_compiled | |
| echo "******************************************" | |
| echo "******* Emb: channel-wise quantized ******" | |
| echo "******************************************" | |
| echo "Running eager" | |
| python generate.py --quant '{"embedding" : {"bitwidth": 8, "groupsize": 0}}' --gguf-path ${GGUF_PATH} --tokenizer-path ${TOKENIZER_PATH} --max-new-tokens 20 --temperature 0 > ./output_eager | |
| cat ./output_eager | |
| echo "Running compiled" | |
| python generate.py --compile --quant '{"embedding" : {"bitwidth": 8, "groupsize": 0}}' --gguf-path ${GGUF_PATH} --tokenizer-path ${TOKENIZER_PATH} --max-new-tokens 20 --temperature 0 > ./output_compiled | |
| cat ./output_compiled | |
| echo "******************************************" | |
| echo "******** Emb: group-wise quantized *******" | |
| echo "******************************************" | |
| echo "Running eager" | |
| python generate.py --quant '{"embedding" : {"bitwidth": 8, "groupsize": 8}}' --gguf-path ${GGUF_PATH} --tokenizer-path ${TOKENIZER_PATH} --max-new-tokens 20 --temperature 0 > ./output_eager | |
| cat ./output_eager | |
| echo "Running compiled" | |
| python generate.py --compile --quant '{"embedding" : {"bitwidth": 8, "groupsize": 8}}' --gguf-path ${GGUF_PATH} --tokenizer-path ${TOKENIZER_PATH} --max-new-tokens 20 --temperature 0 > ./output_compiled | |
| cat ./output_compiled | |
| echo "tests complete" | |
| echo "******************************************" |