version 1.33.0

What's New

• PyTorch
  • Enhancements done in export pipeline for GPU memory optimization with LLMs.
  • [Experimental] Added support for handling of LoRA (via PEFT API) in AIMET. and enabled export of required artifacts for QNN.
  • Added examples for training pipeline with for distributed KD-QAT.
  • [Experimental] Added support for block wise quantization (BQ) to support w4fp16 format, and the low-power block quantization (LPBQ) to support w4a8 and w4a16 formats. This feature needs QuantSim V2.

Documentation

• Release main page: https://github.com/quic/aimet/releases/tag/1.33.0
• Installation guide: https://quic.github.io/aimet-pages/releases/1.33.0/install/index.html
• User guide: https://quic.github.io/aimet-pages/releases/1.33.0/user_guide/index.html
• API documentation: https://quic.github.io/aimet-pages/releases/1.33.0/api_docs/index.html

Packages

• aimet_torch-1.33.0.cu118-cp310-cp310-manylinux_2_34_x86_64.whl
  • PyTorch 2.1 GPU package with Python 3.10 and CUDA 11
• aimet_torch-1.33.0.cu117-cp310-cp310-manylinux_2_34_x86_64.whl
  • PyTorch 1.13 GPU package with Python 3.10 and CUDA 11.x
• aimet_torch-1.33.0.cpu-cp310-cp310-manylinux_2_34_x86_64.whl
  • PyTorch 2.1 CPU package with Python 3.10 - If installing on a machine without CUDA
• aimet_onnx-1.33.0.cu117-cp310-cp310-manylinux_2_34_x86_64.whl
  • ONNX 1.14 GPU package with Python 3.10 - Recommended for use with ONNX models
• aimet_onnx-1.33.0.cpu-cp310-cp310-manylinux_2_34_x86_64.whl
  • ONNX 1.14 CPU package with Python 3.10 - If installing on a machine without CUDA
• aimet_tensorflow-1.33.0.cu118-cp310-cp310-manylinux_2_34_x86_64.whl
  • TensorFlow 2.10 GPU package with Python 3.10 - Recommended for use with TensorFlow models
• aimet_tensorflow-1.33.0.cpu-cp310-cp310-manylinux_2_34_x86_64.whl
  • TensorFlow 2.10 CPU package with Python 3.10 - If installing on a machine without CUDA