intel/auto-round

Pull Requests Commits

feat: add Cosmos3-Super packed quantization scripts

yiliu30 committed 87 days ago

87b413cd

test: cover Cosmos3 diffusion pipeline kwargs

yiliu30 committed 87 days ago

c42d18ad

fix: support Cosmos3 omni diffusion quantization

yiliu30 committed 87 days ago

3ea05ab3

feat: add Cosmos3-Nano quantization support

yiliu30 committed 93 days ago

a4adf972

[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

add the wrapper for woqgemm (#1965)

Zhenzhong1 committed 95 days ago

Verified 3ab95207

[ARK] Add varlen support for sdpa and update tests (#1957)

luoyu-intel committed 96 days ago

Verified 948c7783

Adapt the latest sglang integration test Engine kwargs. (#1964)

changwangss committed 97 days ago

Verified 49244549

fix auto_scheme gguf format wrong bits if use tiny model (#1953)

n1ck-guo committed 97 days ago

Verified b1a6baf9

speed up awq ci (#1954)

WeiweiZhang1 committed 100 days ago

Verified 9468e52c

add other non-integer scheme support for awq (#1892)

WeiweiZhang1 committed 100 days ago

Verified 5a3ade8c

Remove kernels version limit in test (#1945)

chensuyue committed 101 days ago

Verified 142e91b4

Use shared tokenizer save helper in exporters (#1951)

changwangss committed 101 days ago

Verified 436b9db6

update ARK README (#1922)

luoyu-intel committed 101 days ago

Verified f42a0421

Implement shard parallelism configuration and update documentation (#1942)

xin3he committed 102 days ago

Verified 7e2b69d4

fix saving hadamard rotation model without rotation config. (#1933)

lkk12014402 committed 102 days ago

Verified 6a2e5ad3

fix few gguf bugs of hf-name and checkpoint name (#1939)

n1ck-guo committed 102 days ago

Verified 126793d5

Add flashinfer-jit-cache installation to cuda unit test setup (#1947)

XuehaoSun committed 103 days ago

Verified 349b0418

fix issue of step3p7 and CI failure (#1929)

xin3he committed 103 days ago

Verified 1c4baa03

Older