intel/auto-round

Pull Requests Commits

Enhance E8M0 block scale expansion for DeepSeek variants and improve logging for fp32 to UE8M0 conversion

xin3he committed 9 days ago

5c0e8480

Add logging for fp32 to UE8M0 conversion in scale tensor processing

xin3he committed 9 days ago

52adf875

Implement hydration of scale tensors from sibling shards in FP8 dequantization

xin3he committed 9 days ago

1f16c07e

Merge remote-tracking branch 'origin/main' into xinhe/8-4

xin3he committed 9 days ago

fa73368b

fix sparge on torch2.13 (#2130)

yiliu30 committed 9 days ago

Verified d8a30059

Add weight dequantization support and caching mechanism for NVFP4 E5M3

xin3he committed 9 days ago

27cc2654

Refator UT and improve code coverage (#2057)

lvliang-intel committed 9 days ago

Verified 7a75218a

Enhance model conversion and CUDA integration for NVFP4 E5M3 support

xin3he committed 10 days ago

e7e2f015

[fix] XPU crash due to circle call of nonzero() and index_select() (#2117)

xin3he committed 10 days ago

Verified 28003697

[auto_round format] Add NVFP4 E5M3 support with CuTe integration and update related components

xin3he committed 10 days ago

fa58ab62

Add NVFP4 E5M3 support with CuTe QDQ integration and related tests

xin3he committed 10 days ago

814659b1

xin3he committed 10 days ago

89dbe07e

Enable torch.compile for selected tests and update related configurations (#2124)

xin3he committed 10 days ago

Verified a2f6b316

[pre-commit.ci] auto fixes from pre-commit.com hooks

pre-commit-ci[bot] committed 10 days ago

47bfc852

[llm_compressor format] Add NVFP4 E5M3 quantization support and related components

xin3he committed 10 days ago

cc0e5feb

Add NVFP4 E5M3 support and related components for model-free quantization

xin3he committed 10 days ago

cfc5eadb

support fused MoE quantization and export for GGUF format (#2072)

n1ck-guo committed 10 days ago

Verified 1fca0c74

refactor scheme,layer_config and formats (#2013)

n1ck-guo committed 11 days ago

Verified 77e8ea59

Automatically select calibration datasets for code models (#2107)

changwangss committed 11 days ago

Verified f8de00c3

disable parallel for non low_gpu_mem_usage (#2116)

xin3he committed 12 days ago

Verified 417d87de

add opt_rtn support for mxfp4 and nvfp4 (#2100)

wenhuach21 committed 12 days ago

Verified 41cca72f

Fix: temporarily skip tests due to known issues with xfail annotations (#2113)

chensuyue committed 15 days ago

Verified 60b813cb

Enable AutoScheme parallel and disk stream by default (#2105)

xin3he committed 15 days ago

Verified 8a9a1d13

Fix: skip exporting fp8 attention q_max (#2103)

yiliu30 committed 15 days ago

Verified 2a267e72

fix MXFP model free auto-round format bug (#2106)

xin3he committed 15 days ago

Verified 378187c7

[Feature] Register_block_output to support GLM5.2 tuning (#2104)

xin3he committed 15 days ago

Verified d05fb638

feat: add OpenCodeInstruct calibration dataset (#2096)

changwangss committed 16 days ago

Verified ca9524ae

Enable torch.compile by default and update related documentation (#2070)

xin3he committed 16 days ago

Verified be862d37

Stream AutoScheme's per-layer sensitivity scoring block-by-block (#2063)

aquilarubra committed 16 days ago

Verified 73661edb

Fix dataset and batch_size not taking effect in CLI (#2102)

changwangss committed 16 days ago

Verified 212c8bda

Older