intel/auto-round

Pull Requests Commits

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

pre-commit-ci[bot] committed 17 hours ago

1ce1f8eb

Refactor code structure for improved readability and maintainability

xin3he committed 1 day ago

b2926220

Rename fp4_v2_with_global_scale to nvfp4_v2_with_global_scale in nvfp.py and update corresponding tests

xin3he committed 1 day ago

26a36021

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

xin3he committed 1 day ago

caccbeae

Enhance model caching and warnings for NVFP4 E5M3 scheme in model_free.py and model.py; update documentation for llm_compressor support

xin3he committed 1 day ago

fe5e5976

Add support for kimi_k25 INT4 packed source models and related tests (#2156)

xin3he committed 1 day ago

Verified 67437798

Add cpu & xpu nightly test (#2149)

XuehaoSun committed 1 day ago

Verified 436dc247

[ARK] Optimize SYCL-TLA woqgemm_s8 on B70 (#2133)

Zhenzhong1 committed 1 day ago

Verified 2ea60e55

Set the quant_method to auto-round by default (#2138)

Zhenzhong1 committed 1 day ago

Verified abacdadf

Support SVDQuant (#2095)

changwangss committed 1 day ago

Verified ccb36588

refine awq, bugfix (#2115)

WeiweiZhang1 committed 2 days ago

Verified a3af3671

feat: Support world model Cosmos3 (#2040)

lvliang-intel committed 2 days ago

Verified 9181fa95

feat: enable search scale for mxfp in model-free (#2141)

xin3he committed 2 days ago

Verified 9352659f

Add timeout decorators to improve test case stability (#2155)

XuehaoSun committed 2 days ago

Verified cee71092

refine eval, add more option (#2114)

WeiweiZhang1 committed 2 days ago

Verified fe7676de

Migrate cpu sdpa from NS flash attn (#2092)

jijiaz committed 2 days ago

Verified 7402af81

Refactor unit test scripts (#2148)

XuehaoSun committed 2 days ago

Verified 506bd5fa

fix: prevent quantized layers from being reinitialized by Transformers (#2147)

n1ck-guo committed 3 days ago

Verified a5e7393e

update per comments

xin3he committed 4 days ago

dfe6c605

Refactor NVFP4 E5M3 quantization to use "nvfp4_v2" data type

xin3he committed 4 days ago

aa967f87

feat: add per-head fp8 kv attention calibration (#2044)

zeel2104 committed 5 days ago

Verified f17d9cd4

Merge branch 'main' into xinhe/8-4

xin3he committed 5 days ago

Verified c5873505

fix performance regression bug (#2136)

n1ck-guo committed 6 days ago

Verified f8ad63f4

Add NVFP4 quantization support with layer configuration checks in model_free.py for nvidia nvfp4 model reformat

xin3he committed 6 days ago

2b9947ea

Normalize NVFP4 global scales to reciprocal form in _normalize_nvfp4_source_tensors function

xin3he committed 6 days ago

6942d535

Enhance quantization handling by adding NVFP4 support in _get_mxfp_group_scheme_and_format function

xin3he committed 6 days ago

979796f9

Merge branch 'main' into xinhe/8-4

xin3he committed 6 days ago

Verified 2203adf7

Fix tensor reshaping for NVFP4 source tensors in normalization and handling

xin3he committed 6 days ago

639b54d0

Enhance unit tests with timeout handling and stability improvements (#2135)

XuehaoSun committed 6 days ago

Verified e4a02ccd

Add NVFP4 support with normalization and passthrough handling for legacy tensors

xin3he committed 8 days ago

d11676f6

Older