Download Latest Version Version 2.37.0 source code.zip (10.4 MB)
Email in envelope

Get an email when there's a new version of AIMET

Home / 2.36.0
Name Modified Size InfoDownloads / Week
Parent folder
aimet_onnx-2.36.0+cpu-cp310-abi3-win_amd64.whl 2026-07-27 677.4 kB
1 1 weekly downloads
aimet_onnx-2.36.0+cpu-cp310-abi3-manylinux_2_34_x86_64.whl 2026-07-27 778.1 kB
0
aimet_onnx-2.36.0+cpu-cp310-abi3-win_arm64.whl 2026-07-27 647.2 kB
0
aimet_onnx-2.36.0+cu126-cp310-abi3-manylinux_2_34_x86_64.whl 2026-07-27 68.9 MB
0
aimet_torch-2.36.0+cpu-py310-none-any.whl 2026-07-27 792.6 kB
0
aimet_torch-2.36.0+cu126-py310-none-any.whl 2026-07-27 792.6 kB
0
README.md 2026-07-27 1.6 kB
0
Version 2.36.0 source code.tar.gz 2026-07-27 9.6 MB
0
Version 2.36.0 source code.zip 2026-07-27 10.3 MB
0
Totals: 9 Items   92.4 MB 1

2.36.0

  • New Features

    • ONNX
      • Add tensor-level set_precision API to aimet-onnx QuantSim (ad61748)
      • Enable AdaScale for Qwen3.5 (03df10a)
    • Torch
      • Add R2 per-head rotation to aimet-torch SpinQuant (0ba884a)
      • Enable AdaScale for Qwen3.5 (e5f5462)

  • Bug fixes and Improvements

    • ONNX

      • Skip weightless dynamic MatMuls in sequential MSE (8413aea)
      • Fix bugged or missing op types in onnx2torch (6f3255b)
      • Define QcQuantizeOp interface to set precision via qtype (d0c1e4b)
      • Consolidate ONNX decoder block detection into block_topology package (f441f2f)
      • Detect decoder residual writers by graph walk in role map (22e6052)
      • Propagate through float-to-float casts to find effective quantizer (9ca5115)
      • Fix export omitting bias encodings for dynamic convs (326167c)

    • Torch

      • Fall back from Triton to PyTorch if input size exceeds 2^31 (a8636b5)
      • Skip deriving data movement op output encoding for MaskedSoftmax subgraph (75bf0f7)
      • Fall back to chained sampling for cached-region resume in BlockwiseSampler (5341997)
      • Remove erroneous QuantizationMixin.ignore in Gemma4 (dcfe9de)

    • Common

      • Add fp16 support in GenAILab (fec7c2c)
      • Onboard gemma4 qat model to GenAILab (f65cf66)
      • Add SplitFusedLayers adaptation to GenAI Lab (0c1729d)
      • Add MMLU Pro dataset/metric to GenAI Lab (c16100b)
      • Update GenAILab pinned package dependencies (7b3f0fb)
Source: README.md, updated 2026-07-27