intel/auto-round

Pull Requests Commits

update the code fot fix

a32543254 committed 86 days ago

8064d81d

Remove transformers version limit for release package (#2030)

chensuyue committed 87 days ago

Verified 5e586bcb

bug-fix: implement dedicated cache for streaming source shards (#2021)

xin3he committed 87 days ago

Verified ffd3dfe7

fix: CLI ignore --scheme when bits/act_bits are unset (#2019)

n1ck-guo committed 87 days ago

Verified c70e99f6

fix the mismatch between weight and scale in gguf format (#2014)

n1ck-guo committed 87 days ago

Verified b482d150

Fix checkpoint remap (#2001)

yiliu30 committed 87 days ago

Verified ec73a8dd

Add moe prefill/ decode with int2/int4/int8 sym /asym and fp8 e4m3 e5m2 (#1813)

Copilot committed 88 days ago

Verified 74143702

[pre-commit.ci] pre-commit autoupdate (#2004)

pre-commit-ci[bot] committed 89 days ago

Verified 731a4312

[ARK] Support Dense Gemm using sycl-tla (#1968)

Zhenzhong1 committed 89 days ago

Verified 47822429

[ARK] feat: expose smooth_k (use_mean_bias) parameter from Python to C++ in SAGE attention kernels (#1995)

luoyu-intel committed 89 days ago

Verified d55ba004

bump release version to 0.14.0 (#2002)

chensuyue committed 90 days ago

Verified 24bfce00

update rotation doc. (#1999)

lkk12014402 committed 90 days ago

Verified 5ee45db7

pin transformers (#1998)

yiliu30 committed 90 days ago

Verified fcd1c854

Fix cuda ut fail caused by used instructions for base model (#1994)

n1ck-guo committed 90 days ago

Verified 43f443e6

feat: add return_lse support for SDPA and SAGEV1 kernel — enable sequence-parallel (Ring Attention) merge (#1991)

luoyu-intel committed 92 days ago

Verified b481cebb

Enable auto-round-bot for CI comments (#1993)

chensuyue committed 93 days ago

Verified b63a89f5

Implement AutoScheme support for model-free packing with validation a… (#1990)

xin3he committed 93 days ago

Verified cd10607c

Update CUDA UT to cu130 (#1992)

XuehaoSun committed 93 days ago

Verified 52564823

[AutoScheme] fix lm_head no_grad issue and support MOE model with shared mix_score (#1971)

xin3he committed 93 days ago

Verified e6b35d13

fix: use spawn instead of fork for dataset subprocess on macOS (#1980)

ramkrishs committed 94 days ago

Verified 1a7ed495

fix few bugs cause less accuracy of gguf format (#1960)

n1ck-guo committed 94 days ago

Verified e4404be7

Minor fix for auto-round-lib CI workflow (#1983)

chensuyue committed 94 days ago

Verified c945a182

fix: promote None-initialized kwarg to list in forward_capture (#1979)

ramkrishs committed 94 days ago

Verified 0622d472

fix: avoid silent weight loss when OffloadManager fails to save a block (#1982)

xin3he committed 94 days ago

Verified 64378412

rename the base rotation (deterministic/random hadamard) directory name. (#1956)

lkk12014402 committed 94 days ago

Verified 45dcb198

Add llmc ut back (#1974)

yiliu30 committed 95 days ago

Verified d763ae04

Fix gguf cuda ut (#1975)

n1ck-guo committed 95 days ago

Verified 1c6f4e96

Fix mlx test path (#1952)

chensuyue committed 95 days ago

Verified 911cbe12

Fix layer_config bug and CI unstable issue (#1949)

xin3he committed 95 days ago

Verified 05e67950

Skip fused_moe replacement when llm-compressor oneshot already linearizes experts (#1966)

changwangss committed 95 days ago

Verified 09d3dec1

Older