| ############################################################################### | |
| echo == Prep Enviroment == | |
| # Pull both the model folder and llama.cpp (for the conversion script) | |
| #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 == Convert from safetensor to gguf == | |
| ./llama.cpp/convert.py maykeye_tinyllama --outtype f16 --metadata maykeye_tinyllama-metadata.json | |
| mv maykeye_tinyllama/TinyLLama-v0-5M-F16.gguf TinyLLama-v0-5M-F16.gguf | |
| echo == Generating Llamafile == | |
| cp ./llamafile/o/llama.cpp/main/main TinyLLama-v0-5M-F16.llamafile | |
| # Create an .args file with settings defaults | |
| cat >.args <<EOF | |
| -m | |
| TinyLLama-v0-5M-F16.gguf | |
| EOF | |
| # zip align engine, gguf and default args | |
| zipalign -j0 \ | |
| TinyLLama-v0-5M-F16.llamafile \ | |
| TinyLLama-v0-5M-F16.gguf \ | |
| .args | |
| ############################################################################### | |
| echo == Test Output == | |
| ./TinyLLama-v0-5M-F16.llamafile --cli -p "hello world the gruff man said" | |