| MODEL_DIR="maykeye_tinyllama" | |
| METADATA_FILE="maykeye_tinyllama-metadata.json" | |
| ############################################################################### | |
| # Pull both model folder, llamafile (for the engine) and llama.cpp (for the conversion script) | |
| echo == Prep Enviroment == | |
| #git submodule update --init | |
| ############################################################################### | |
| echo == Build and prep the llamafile engine execuable == | |
| pushd llamafile | |
| make -j8 | |
| make | |
| # This is where each executables is located for reference purpose for now as of 2024-04-05 | |
| # and was determined by running `sudo make install PREFIX=/usr/local` | |
| # ./o/llamafile/zipalign --> /usr/local/bin/zipalign | |
| # ./o/llama.cpp/main/main --> /usr/local/bin/llamafile | |
| # ./o/llama.cpp/imatrix/imatrix --> /usr/local/bin/llamafile-imatrix | |
| # ./o/llama.cpp/quantize/quantize --> /usr/local/bin/llamafile-quantize | |
| # ./build/llamafile-convert --> /usr/local/bin/llamafile-convert | |
| # ./o/llama.cpp/perplexity/perplexity --> /usr/local/bin/llamafile-perplexity | |
| # ./o/llama.cpp/llava/llava-quantize --> /usr/local/bin/llava-quantize | |
| popd | |
| ############################################################################### | |
| echo == What is our llamafile name going to be? == | |
| ./llama.cpp/convert_hf_to_gguf.py ${MODEL_DIR} --metadata ${METADATA_FILE} --outtype f16 --dry-run 2>/dev/null | |
| OUTFILE_PATH=$(./llama.cpp/convert_hf_to_gguf.py ${MODEL_DIR} --metadata ${METADATA_FILE} --outtype f16 --dry-run 2>/dev/null) | |
| OUTFILE_FILE=$(basename ${OUTFILE_PATH}) | |
| OUTFILE="${OUTFILE_FILE%.gguf}" | |
| echo We will be aiming to generate $OUTFILE.llamafile | |
| ############################################################################### | |
| echo == Convert from safetensor to gguf == | |
| ./llama.cpp/convert_hf_to_gguf.py ${MODEL_DIR} --metadata ${METADATA_FILE} --outtype f16 --verbose | |
| mv ${MODEL_DIR}/${OUTFILE}.gguf ${OUTFILE}.gguf | |
| # Generate Diagnostics Dumpfile | |
| ./llama.cpp/gguf-py/scripts/gguf_dump.py --markdown ${OUTFILE}.gguf > ${OUTFILE}.dump.md | |
| ############################################################################### | |
| echo == Generating Llamafile == | |
| cp ./llamafile/o/llama.cpp/main/main ${OUTFILE}.llamafile | |
| # Create an .args file with settings defaults | |
| cat >.args <<EOF | |
| -m | |
| ${OUTFILE}.gguf | |
| EOF | |
| # zip align engine, gguf and default args | |
| ./llamafile/o/llamafile/zipalign -j0 ${OUTFILE}.llamafile ${OUTFILE}.gguf .args | |
| ############################################################################### | |
| echo == Test Output ./${OUTFILE}.llamafile == | |
| ./${OUTFILE}.llamafile --cli -p "hello world the gruff man said" | |