Continuously optimize AutoScheme RAM consumption#1703

Open

lvliang-intel wants to merge 45 commits into

mainfrom

lvl/autoscheme_ram_opt

Commits on Apr 17, 2026

Commits on May 12, 2026

fix: add missing run_mllm entry point alias (#1695)

JGSphaela
authored and
lvliang-intel
committed
rename scheme INT8_W8A8 to INT8 (#1687)

authored and
lvliang-intel
committed
update mtp quant for special cases (#1691)

authored and
lvliang-intel
committed
Update gaudi-docker to v1.24.0 & fix CUDA UT (#1708)

XuehaoSun
authored and
lvliang-intel
committed
add support for gemma4 model (#1655)

committed
ignore mtp.fc for qwen3_5 due to vllm failure (#1710)

xin3he
authored and
lvliang-intel
committed
[research feature] Introduce INT4 support at the algorithm level (#1641)

authored and
lvliang-intel
committed
refine int4 doc (#1720)

authored and
lvliang-intel
committed
Support new model Qwen/Qwen3.6-35B-A3B (#1705)
lvliang-intel
committed
Revert "ignore mtp.fc for qwen3_5 due to vllm failure (#1710)" (#1730)

xin3he
authored and
lvliang-intel
committed
skip quantizing mtp.fc since vLLM doesn't support (#1731)

xin3he
authored and
lvliang-intel
committed
Update pull_request_template.md (#1727)

xin3he
authored and
lvliang-intel
committed
Create model_support_request.yml (#1738)

xin3he
authored and
lvliang-intel
committed
fix gemma3 gguf ut fail (#1735)

n1ck-guo
authored and
lvliang-intel
committed
Remove threaded packing from exporters (#1719)

authored and
lvliang-intel
committed
add small zimage test and fix bug (#1734)

xin3he
authored and
lvliang-intel
committed
Enhance llmc CI on XPU (#1483)

authored and
lvliang-intel
committed
Reduce xpu memory usage with patch_xpu_sdpa_drop_causal_mask (#1716)

authored and
lvliang-intel
committed
[Experimental]Add MLX format export support and AutoScheme for vlm support (#1732)

authored and
lvliang-intel
committed
add warnings for lm_head activation scale fallback (#1728)

n1ck-guo
authored and
lvliang-intel
committed
add support for MiMo-V2-Flash (#1718)

n1ck-guo
authored and
lvliang-intel
committed
New architecture for auto_round (#1542)

n1ck-guo
authored and
lvliang-intel
committed
Fix vllm CUDA CI (#1750)

XuehaoSun
authored and
lvliang-intel
committed
delete unreproduced results for now (#1760)

ZaneMark
authored and
lvliang-intel
committed
Fix hpu error (#1766)

authored and
lvliang-intel
committed
[MTP]split gate_up_proj and fix accu gap in rtn quantization (#1758)

xin3he
authored and
lvliang-intel
committed
clean and fix for new arch (#1761)

authored and
lvliang-intel
committed
support gptqmodel 7.0.0 and fix bug in CI (#1772)

authored and
lvliang-intel
committed
Optimize CUDA CI and Code Scan workflows (#1770)

authored and
lvliang-intel
committed
fix accuracy regression and check it in CUDA CI (#1785)

authored and
lvliang-intel
committed
fix amp (#1768)

wenhuach21
authored and
lvliang-intel
committed
fix amp (#1767)

authored and
lvliang-intel
committed
Fix incompatible weight names (#1759)

authored and
lvliang-intel
committed
add notes (#1795)

wenhuach21
authored and
lvliang-intel
committed
remove IPEX related code, doc, and test (#1787)

xin3he
authored and
lvliang-intel
committed
support model_free WOQ quantization (#1699)

authored and
lvliang-intel
committed
Integrate AutoRound Lib (#1723)

Zhenzhong1
authored and
lvliang-intel
committed
fix new arch bug for llmc (#1781)

n1ck-guo
authored and
lvliang-intel
committed
fix bug of gguf alg ext (#1796)

n1ck-guo
authored and
lvliang-intel
committed
Continue optimizing AutoScheme RAM consumption
lvliang-intel
committed
[pre-commit.ci] auto fixes from pre-commit.com hooks

pre-commit-ci[bot]
authored and
lvliang-intel
committed
Merge branch 'main' into lvl/autoscheme_ram_opt
lvliang-intel
authored
Merge branch 'main' into lvl/autoscheme_ram_opt
lvliang-intel
authored