Pull Requests intel/auto-round

feat: export SVDQuant NVFP4 checkpoints for vLLM-Omni

#2445 opened 2026-10-08 03:10 by changwangss

fix: give attn_v of LLM_TYPE_70B models Q5_K like llama.cpp

#2443 opened 2026-10-08 01:57 by raashish1601 0.16.1

Add requests into main requirements.txt

#2441 opened 2026-10-08 01:51 by chensuyue 0.16.1

fix: keep make_q3_quants sums in sync with accepted levels

#2439 opened 2026-10-07 22:26 by raashish1601 0.16.1

fix: keep 0 in the range of GGUF asymmetric K-quant groups

#2438 opened 2026-10-07 22:06 by raashish1601 0.16.1

fix: correct Q5_K GGUF packing

#2437 opened 2026-10-07 21:51 by raashish1601 0.16.1

fix: match layer names ending with a digit in ignore_layers

#2435 opened 2026-10-01 16:09 by kwy404 0.16.1

fix: preserve calibration dataset on memory errors

#2433 opened 2026-09-30 09:48 by Shubham-Padkonde

fix: preserve GLM nextn weights during checkpoint copying

#2430 opened 2026-09-30 08:02 by Shubham-Padkonde

[feat] support deepseek_v41: switch blockwise mxfp8 to OCP mxfp8.

#2426 opened 2026-09-29 13:36 by lkk12014402

fix(multi-gpu): disable torch.compile on multi-device execution to prevent ContextVar crash

#2422 opened 2026-09-24 13:56 by reginaldalfret

fix: use spawn start method for vLLM evaluation to avoid CUDA re-initialization error

#2421 opened 2026-09-24 11:45 by reginaldalfret

fix: keep escapes and quantified dots in to_standard_regex

#2418 opened 2026-09-24 08:21 by Shubham-Padkonde

refactor: move diffusion tuning data handling out of SignRound

#2404 opened 2026-09-21 11:06 by changwangss

Reset context singletons between tests

#2390 opened 2026-09-20 07:36 by Shubham-Padkonde

Don't rewrite packed modules as float weights when resuming

#2386 opened 2026-09-20 02:54 by Shubham-Padkonde

feat: chunked tuning for lm_head-class layers

#2383 opened 2026-09-20 00:23 by avtc

feat: add mxfp8 mxfp4 moe bdpas prefill

#2371 opened 2026-09-16 17:21 by a32543254

[bugfix][refine]: drop no-op RotationConfig.__init__

#2365 opened 2026-09-15 08:56 by lkk12014402

refine dataset

#2363 opened 2026-09-15 08:46 by wenhuach21

feat: multi-GPU quantization improvements -- batched device-local sea…

#2357 opened 2026-09-14 18:37 by avtc 0.17.0

feat: data-parallel block tuning via --parallel_quantization

#2351 opened 2026-09-11 19:09 by avtc

Refine device operations into a unified DeviceManager and ARDevice abstraction

#2344 opened 2026-09-11 07:32 by lvliang-intel 0.17.0

feat(ark): add INT4 S4 pre-packed Q*K kernel for SageAttention

#2319 opened 2026-09-08 07:03 by luoyu-intel

Enhance offload cleanup handling for exception cases

#2317 opened 2026-09-08 01:53 by lvliang-intel

Recurrent Residual Quantization (RRQ) for LLMs

#2308 opened 2026-09-06 14:00 by luoyu-intel

support teq algo WIP experimental

#2301 opened 2026-09-04 06:32 by WeiweiZhang1

Xpu hmt mxfp4

#2264 opened 2026-08-31 12:09 by a32543254

Add lagrangian solver in AutoScheme

#2221 opened 2026-08-24 13:25 by wenhuach21

feat: W4A8 ARK XPU MoE kernel (int4 weight / int8 compute) with prefill + decode

#2143 opened 2026-08-11 05:06 by Copilot