amdsharktank 3.9.0

amdshark layers and inference models for genai

amdshark Tank

WARNING: This is an early preview that is in progress. It is not ready for general use.

Light weight inference optimized layers and models for popular genai applications.

This sub-project is a work in progress. It is intended to be a repository of layers, model recipes, and conversion tools from popular LLM quantization tooling.

Project Status

Examples

The repository will ultimately grow a curated set of models and tools for constructing them, but for the moment, it largely contains some CLI examples. These are all under active development and should not yet be expected to work.

Perform batched inference in PyTorch on a paged llama derived LLM:

Note: Use --device='cuda:0' to run this inference on an AMD GPU.

python -m amdsharktank.examples.paged_llm_v1 \
  --hf-dataset=open_llama_3b_v2_f16_gguf \
  --device='cuda:0' \
  "Prompt 1" \
  "Prompt 2" ...

Export an IREE compilable batched LLM for serving:

python -m amdsharktank.examples.export_paged_llm_v1 \
  --hf-dataset=open_llama_3b_v2_f16_gguf \
  --output-mlir=/tmp/open_llama_3b_v2_f16.mlir \
  --output-config=/tmp/open_llama_3b_v2_f16.json

Generate sample input tokens for IREE inference/tracy:Add commentMore actions

python -m amdsharktank.examples.paged_llm_v1 \
  --irpa-file=open_llama_3b_v2_f16.irpa \
  --tokenizer-config-json=tokenizer_config.json \
  --prompt-seq-len=128 \
  --bs=4 \
  --dump-decode-steps=1 \
  --max-decode-steps=1 \
  --dump-path='/tmp' \
  --device='cuda:0'

Dump parsed information about a model from a gguf file:

python -m amdsharktank.tools.dump_gguf --hf-dataset=open_llama_3b_v2_f16_gguf

Package Python Release Builds

• To build wheels for Linux:

  ./build_tools/build_linux_package.sh
  

  That should produce build_tools/wheelhouse/amdsharktank-{X.Y.Z}.dev0-py3-none-any.whl, which can then be installed with

  python3 -m pip install build_tools/wheelhouse/amdsharktank-{X.Y.Z}.dev0-py3-none-any.whl
  

• To build a wheel for your host OS/arch manually:

  # Build amdsharktank.*.whl into the dist/ directory
  #   e.g. `amdsharktank-3.0.0.dev0-py3-none-any.whl`
  python3 -m pip wheel -v -w dist .
  
  # Install the built wheel.
  python3 -m pip install dist/*.whl