intel/auto-round

Pull Requests Commits

set config for mixed-precision model backend initialization

xin3he committed 26 days ago

55a27976

use AutoModelForImageTextToText instead of deprecated AutoModelForVision2Seq

xin3he committed 27 days ago

f00c584d

add try except for 2bits evaluation

xin3he committed 27 days ago

aab0b456

Support quant for `meituan-longcat/LongCat-Flash-Lite` (#1388)

Kaihui-intel committed 27 days ago

Verified a12e3f5a

Fix version (#1393)

XuehaoSun committed 27 days ago

Verified 68d07bfa

Implement MXFP4Handler MXFP8Handler NVFP4Handler (#1385)

xin3he committed 27 days ago

Verified f7db867a

fix layer config (#1373)

wenhuach21 committed 27 days ago

Verified 4064c4f1

[Regression] Detach scale tensor to prevent holding computation graph references (#1389)

xin3he committed 28 days ago

Verified 1c758f45

fix cuda ut fail (#1370)

n1ck-guo committed 28 days ago

Verified 6cdef7a2

Support general MOE replacement for MOE models (Transformers 5.0 compatible) (#1374)

lvliang-intel committed 28 days ago

Verified 28839a9a

Disable replace `FP8Expert` (#1379)

yiliu30 committed 28 days ago

Verified 137da551

fix --device_map cuda and xpu issue (#1383)

wenhuach21 committed 28 days ago

Verified f132c0c4

rm duplicate args of the quantization extra config (#1334)

WeiweiZhang1 committed 28 days ago

Verified 7e4ba88a

fix exporting fp8 static for format auto_round (#1381)

n1ck-guo committed 29 days ago

Verified 93f7703a

Standardize comment language in utils.py (#1378)

shantoislamdev committed 29 days ago

Verified 108f2b58

Add Links; Add step_by_step_CN; Fix several grammatical issues (#1366)

ZaneMark committed 29 days ago

Verified 3b7b87ec

Extensible Module Weight Type Conversion Framework. (#1375)

xin3he committed 29 days ago

Verified 53f20cf1

Fix FP8 MLA (#1343)

yiliu30 committed 29 days ago

Verified db77f3b7

feat: handle per-tensor FP8 dequantization for Devstral models (#1356)

SwekeR-463 committed 30 days ago

Verified eddb6de2

fix copy python file issue (#1372)

wenhuach21 committed 32 days ago

Verified 29956d91

add support for w4a16_mixed (#1326)

n1ck-guo committed 32 days ago

Verified 7cff6af5

fix cuda ut fail for transformers v5.0 (#1357)

n1ck-guo committed 32 days ago

Verified eb2ccdc3

support qwen3-next/glm4-moe/deepseekv3 monkey patch for transformers 5 (#1367)

wenhuach21 committed 32 days ago

Verified 0c0a0c1d

support qwen3-moe/glm-4.7-flash monkey patch for transformers 5 (#1354)

wenhuach21 committed 33 days ago

Verified 6bc1d344

fix bug of exporting fp8 static (#1361)

n1ck-guo committed 33 days ago

Verified b04f69c4

Refactor FP8 dequantization and detection using registry pattern (#1348)

scopophobic committed 33 days ago

Verified eaabde4a

refactor auto_round api, using local symbol table to pass all args (#1358)

n1ck-guo committed 33 days ago

Verified 528c9c82

Autoround in vLLM Office Hours (#1322)

yiliu30 committed 34 days ago

Verified b3f6f1fb

Fix cpu ut for transformers v5 (#1333)

Kaihui-intel committed 34 days ago

Verified f6e7e303

Fix CPU CI (#1344)

XuehaoSun committed 34 days ago

Verified 8eec5d8a

Older