intel/auto-round

Pull Requests Commits

package native MiniMax H3 W4A16 FL2VA

yiliu30 committed 23 days ago

9e702c3c

test: add MiniMax-H3 packed MXFP8 Diffusers smoke

yiliu30 committed 27 days ago

2f747bca

feat: support MiniMax-H3 modular diffusion handoff

yiliu30 committed 27 days ago

789721d9

Add random-init diffusion quantization path

Yi Liu committed 34 days ago

9b02b5c3

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

xin3he committed 34 days ago

Verified a2f6b316

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

n1ck-guo committed 35 days ago

Verified 1fca0c74

refactor scheme,layer_config and formats (#2013)

n1ck-guo committed 35 days ago

Verified 77e8ea59

Automatically select calibration datasets for code models (#2107)

changwangss committed 36 days ago

Verified f8de00c3

disable parallel for non low_gpu_mem_usage (#2116)

xin3he committed 36 days ago

Verified 417d87de

add opt_rtn support for mxfp4 and nvfp4 (#2100)

wenhuach21 committed 37 days ago

Verified 41cca72f

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

chensuyue committed 39 days ago

Verified 60b813cb

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

xin3he committed 40 days ago

Verified 8a9a1d13

Fix: skip exporting fp8 attention q_max (#2103)

yiliu30 committed 40 days ago

Verified 2a267e72

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

xin3he committed 40 days ago

Verified 378187c7

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

xin3he committed 40 days ago

Verified d05fb638

feat: add OpenCodeInstruct calibration dataset (#2096)

changwangss committed 40 days ago

Verified ca9524ae

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

xin3he committed 40 days ago

Verified be862d37

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

aquilarubra committed 41 days ago

Verified 73661edb

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

changwangss committed 41 days ago

Verified 212c8bda

Remove RoutedExperts since the bug is fixed. (#2098)

xin3he committed 42 days ago

Verified 8b0923c8

refine alg and support lfq (#2082)

wenhuach21 committed 42 days ago

Verified f1af646b

ark: add SpargeAttn support (#2027)

yiliu30 committed 42 days ago

Verified 5ce9733c

Decoupling oneDNN (#2049)

Zhenzhong1 committed 43 days ago

Verified 54b4b84c

[Feature]: speedup autoscheme with multi-process and cache files (#2083)

xin3he committed 43 days ago

Verified 288f848b

fix gguf format gemma4 High RAM and glm-4.7 error (#2084)

n1ck-guo committed 43 days ago

Verified 0212c163

Temporarily skip the issue in the CT (#2093)

xin3he committed 43 days ago

Verified 33ec4c75

Fix Flux diffusion state propagation during block replay (#2086)

changwangss committed 46 days ago

Verified 3cae4b68

[high-risk][Step 1]Refactor compressor and quantizer (#2039)

wenhuach21 committed 48 days ago

Verified 8f8c8e9d

Add CUDA CI coverage (#2074)

XuehaoSun committed 48 days ago

Verified 6ff414b1

fix: AutoScheme embedding budget and gguf mixed-bits config for multimodal models (#2055)

n1ck-guo committed 51 days ago

Verified 06c33eae

Older