intel/auto-round

Pull Requests Commits

Add per-WrapperLinear activation checkpointing for memory reduction

yiliu30 committed 103 days ago

bbfb992c

Fix MoE for llama/gpt-oss (#1557)

yiliu30 committed 103 days ago

Verified 555052e7

fix missing change for compress_layer_names and add UTs (#1556)

xin3he committed 103 days ago

Verified 6d17bbfb

Fix shard finalization issue to prevent skipping non-LLM tail layers (#1548)

lvliang-intel committed 104 days ago

Verified 81e79a14

Optimize mtp tensor process by compiling layer names once for all (#1554)

xin3he committed 104 days ago

Verified 1111bcd4

Use AR_WORK_SPACE/offload/ as base dir for offload temp files (#1552)

yiliu30 committed 104 days ago

Verified 803912aa

fix wrong match of ignore_layers and use warning instead of error for mismatch (#1553)

xin3he committed 104 days ago

Verified d8ee85c8

Support block-wise fp8 quant (#1487)

mengniwang95 committed 104 days ago

Verified e6600591

fix google/gemma-3-4b-it (#1547)

xin3he committed 105 days ago

Verified b5fc1724

Add markdown-link-check (#1531)

XuehaoSun committed 105 days ago

Verified 50c4c7f4

fix alg_ext torch compile issue (#1540)

wenhuach21 committed 108 days ago

Verified 9d54ddcc

Fix CI test_transformers issue caused by transformers 5.3.0 (#1535)

lvliang-intel committed 108 days ago

Verified 7b087fb2

Fix sglang ut (#1541)

mengniwang95 committed 108 days ago

Verified 53f2b0e0

support MTP module WOQ quantization and step3p5 model quantization (#1526)

xin3he committed 109 days ago

Verified 42fdb069

fix missing trust_remote_code in unfused_moe (#1532)

xin3he committed 110 days ago

Verified 837888dd

fix dynamic int8 w8a8 export issue with tuning (#1525)

thuang6 committed 110 days ago

Verified a42fd85d

fix broken link (#1530)

xin3he committed 110 days ago

Verified f8e8ae9e

Optimize CPU RAM peak memory during quantization (#1386)

lvliang-intel committed 111 days ago

Verified 0d372151

Support diffusion model saving (#1519)

mengniwang95 committed 111 days ago

Verified a0ad7780

support gpt-oss mxfp4 directly loading (#1401)

xin3he committed 111 days ago

Verified e14e7db6

enable --eval for diffusion model (#1522)

xin3he committed 111 days ago

Verified eec1bed6

Fix transformers 5.2.0 transpose conversion issue (#1475)

lvliang-intel committed 111 days ago

Verified 0127590c

Refactor build process to use 'uv build' instead of 'python setup.py' (#1495)

XuehaoSun committed 111 days ago

Verified 50246114

Fix Inference tensors do not track version counter error (#1502)

mengniwang95 committed 111 days ago

Verified e2b88a5c

Enable more xpu instance for xpu CI (#1521)

chensuyue committed 111 days ago

Verified 30e4043f

fix bug of quantizing Z-image (#1516)

xin3he committed 111 days ago

Verified 8a404ec0

Fix CUDA UT and HPU UT (#1514)

XuehaoSun committed 111 days ago

Verified 89f99ae0

release CUDA memory in WeightConverter and avoid meaningless print (#1498)

xin3he committed 111 days ago

Verified cd10eab0

fix low_gpu default value, refine doc (#1520)

WeiweiZhang1 committed 112 days ago

Verified 279637de

[BUG] update moe check logic and make .gate ignore general (#1517)

xin3he committed 112 days ago

Verified 2eaa3b14

Older