intel/auto-round

Pull Requests Commits

Merge branch 'main' into feat/rrq-phase1

luoyu-intel committed 1 hour ago

Verified d99b2a26

fix: always save key AutoRound hyper-parameters to config.json (#2597)

n1ck-guo committed 1 hour ago

Verified c09906f5

Improve CI UT coverage (#2560)

chensuyue committed 1 hour ago

Verified db3a3ee1

fix: use one-dimensional masks for concatenated calibration data (#2607)

changwangss committed 1 hour ago

Verified 520f5d4c

Fix: ARK `return_lse` returns identical LSE for every 16-query-row block (and wrong LSE under `smooth_k=True`) (#2578)

luoyu-intel committed 5 hours ago

Verified ee6e601d

feat: add MXFP4 export to SVDQuant Omni format (#2535)

yiliu30 committed 5 hours ago

Verified edf0d278

Merge branch 'main' into feat/rrq-phase1

luoyu-intel committed 5 hours ago

Verified 1dfbda6d

chore: remove stale backends/rrq.py file still in the index

luoyu-intel committed 5 hours ago

4dfe2324

fix: reset BOS/EOS bookkeeping per sample when packing calibration data (#2591)

urakozz committed 22 hours ago

Verified b3d11504

fix: update test_quant_lm_head to use tiny untied Qwen model path and assert quantization config (#2604)

chensuyue committed 1 day ago

Verified c9c3ce88

test: include auto_round:rrq in public-format contract set

luoyu-intel committed 1 day ago

f9ab7def

Enhance model-free quantization with improved error handling and shard management (#2566)

xin3he committed 1 day ago

Verified 2679ca9a

[fix] use BF16 KV for reference output; use QDQ NVFP4 KV for tuning inference. (#2592)

xin3he committed 1 day ago

Verified a6efc3e4

Update calibration_acc.md (#2599)

ZaneMark committed 3 days ago

Verified 34d16e91

[Examples] Add recipe notebooks for Qwen3.5-4B WNA16 mixed-precision and MiMo-V2.6-Distill-9B (#2589)

vishvaRam committed 3 days ago

Verified d656a7fc

fix: add inplace param to save_rrq_base_model to match save_checkpoint delegation

luoyu-intel committed 3 days ago

7985c1b0

merge: resolve conflicts from main (PR #2308)

luoyu-intel committed 3 days ago

ef130026

fix: restore torch.compile for NVFP4 quantization (#2582)

WeiweiZhang1 committed 3 days ago

Verified fe9a83cd

feat: share one scheme across vLLM-fused projections in AutoScheme by default (#2542)

lvliang-intel committed 4 days ago

Verified 0ea2c3f3

refine export and format code and provide unify api (#2540)

n1ck-guo committed 4 days ago

Verified caf6388d

fix: default AWQ to optimized RTN (disable_opt_rtn=False) (#2572)

WeiweiZhang1 committed 4 days ago

Verified 6c6566f8

fix: support Nemotron-H Omni without mamba_ssm (#2537)

urakozz committed 4 days ago

Verified e73b38a4

fix: support Diffusers weight filenames in low-memory mode (#2580)

changwangss committed 5 days ago

Verified 9c17113a

[High Risk] Support Tensor Parallelism and LoRA, Optimize FP4 MoE Tuning Speed, and Refactor the Code (#2567)

wenhuach21 committed 5 days ago

Verified 5db10829

support for recipes such as ar-best ar-light (#2547)

n1ck-guo committed 6 days ago

Verified 8869842f

fix: don't let MLLM save_quantized default override resolved format

chensuyue committed 6 days ago

63161b66

fix(rrq): resolve ruff F821 + RUF013 pre-commit failures

luoyu-intel committed 6 days ago

072912b2

fix(rrq): accept enable_neuqi in RRQConfig constructor

luoyu-intel committed 6 days ago

7296e738

docs(rrq): move RRQ entry to 2026/10 in What's New

luoyu-intel committed 6 days ago

6024cbe7

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

pre-commit-ci[bot] committed 6 days ago

0a501b95

Older