intel/auto-round

Pull Requests Commits

feat: Add map_max_scale_to_power2 function for mapping linear max_scale to power-of-two ladder

xin3he committed 57 days ago

d22e4838

fix: Update minmax_scale_bound to (0.0, 2.0)

xin3he committed 57 days ago

20c47ab1

change LinearLR to CosineAnnealingLR

xin3he committed 57 days ago

7c117059

feat: Implement compute_mx_v_scale function for per-element v step-scale in MXFP quantization

xin3he committed 57 days ago

0694f9dd

fix: kimi 2.6 dequantization is not correct (#2260)

xin3he committed 57 days ago

Verified 1ea1d3cd

feat: Add support for deterministic algorithms flags in CLI and update logging behavior (#2261)

xin3he committed 57 days ago

Verified d72933bc

Bug fix: Replace WAWrapperLinear with FakeActQuantLinear instead of Linear (#2245)

xin3he committed 57 days ago

Verified 6b29e9c7

fix: warm up Triton CUDA driver before mocking device capability (#2253)

XuehaoSun committed 57 days ago

Verified a97b91a2

bump release version (#2257)

chensuyue committed 59 days ago

Verified dccf193d

disable torch compile in rtn/opt-rtn/iters<10 and fix glm-5.3 tuning exception (#2246)

wenhuach21 committed 60 days ago

Verified 4dbbd15d

[Bugfix] Fix Sglang e2e UT issue (#2250)

lvliang-intel committed 60 days ago

Verified d77c9eaa

fix argument mismatch (#2248)

xin3he committed 61 days ago

Verified 53f3802a

fix: preserve diffusion mixed dtypes during loading (#2232)

changwangss committed 61 days ago

Verified 7585f13c

fix: remove stale quantizer access in dual-transformer diffusion (#2236)

changwangss committed 61 days ago

Verified 68f8aa32

Update troubleshooting tips for quantization issues in README_CN.md (#2244)

XuehaoSun committed 61 days ago

Verified 7eb007ad

test: skip HPU FP8 test with transformers 5.16.x (#2247)

changwangss committed 61 days ago

Verified c3a74f30

increase precision of loss logging in quantization tuning for debugging (#2229)

xin3he committed 61 days ago

Verified 385a48c9

test: skip FP8 tests with transformers 5.16.x (#2241)

changwangss committed 62 days ago

Verified b4628f15

Fix CUDA UT issue (#2238)

xin3he committed 62 days ago

Verified 9e030fd1

Fix Cuda UT coverage collect (#2234)

chensuyue committed 62 days ago

Verified e5693be8

refactor uts, remove duplicate uts and build common uts for all devices to increase coverage (#2157)

n1ck-guo committed 63 days ago

Verified 6410bc20

fix: AutoScheme parallel scoring OOMs (MoE memory, VLM gate) and disk-stream tuning crash (#2220)

avtc committed 63 days ago

Verified 7da9bea8

Fix FP8 quantization config for native fp8 export (#2228)

Zhenzhong1 committed 63 days ago

Verified 017adc89

add rotation ut for xpu. (#2094)

lkk12014402 committed 63 days ago

Verified 60285730

add enable_torch_compile mark to SGLang and vLLM throughput tests (#2226)

xin3he committed 63 days ago

Verified 4b7cc7e8

Update alg combination news (#2227)

wenhuach21 committed 63 days ago

Verified bd9a13ce

update auto_scheme_acc doc (#2000)

n1ck-guo committed 64 days ago

Verified 5683ea15

add awq torch_compile fix, disable_opt_rtn by default for awq (#2208)

WeiweiZhang1 committed 64 days ago

Verified 611d4402

disable timeout in pytest until found a more stable method (#2222)

chensuyue committed 64 days ago

Verified 96ce4480

Preserve diffusion component dtypes during quantization (#2205)

changwangss committed 64 days ago

Verified d6ead479

Older