v0.9.2 patch release

Remove accelerate version limitation #1090

v0.9.1 patch release

Fix installation on ARM devices.

v0.9.0

Highlights

• support automatic mixed bits assignment by @wenhuach21 in #851
• optimize rtn for int woq by @wenhuach21 in #924
• support for model scope by @n1ck-guo in #957
• enhance auto device map and support XPU by @xin3he in #961
• support for immediate saving to reduce ram usage by @Kaihui-intel in #965
• update gguf alg ext by @n1ck-guo in #1026

What's Changed

• Fix rtn tuning_device issue by @Kaihui-intel in #893
• fix vlm gguf ut by @n1ck-guo in #895
• update alg_ext.abi3.so with python compatible version by @chensuyue in #894
• move ste from quant to round for nvfp4 by @xin3he in #889
• Add GPT-OSS quant support by @yiliu30 in #887
• better help printing information by @n1ck-guo in #883
• speedup quant and evaluation, fix recompile issue by @xin3he in #897
• fix nvfp act quantization bug by @WeiweiZhang1 in #891
• support automatic mixed bits assignment by @wenhuach21 in #851
• try to fix gguf vram issue on windows by @wenhuach21 in #886
• remove numba from requirments by @yiliu30 in #905
• Extend mxfp loading dtypes by @yiliu30 in #907
• block dataset logger info by @n1ck-guo in #908
• fix torch compile issue in AutoScheme by @wenhuach21 in #909
• Revert "Extend mxfp loading dtypes" by @wenhuach21 in #915
• support disable_opt_rtn in auto-scheme by @wenhuach21 in #913
• fix llama 4 ut by @n1ck-guo in #896
• Add numba for cpu lib by @yiliu30 in #919
• Loosen the packing restrictions for mxfp&nvfp by @WeiweiZhang1 in #911
• Extend mxfp loading dtypes by @yiliu30 in #916
• Fix act config exporting for mixed schemes by @WeiweiZhang1 in #903
• optimize rtn for int woq by @wenhuach21 in #924
• fix bug of gguf and support for LiquidAI/LFM2-1.2B by @n1ck-guo in #927
• remove numpy<2.0 limitation by @xin3he in #921
• enable regex quantization config saving for mixed bits by @WeiweiZhang1 in #825
• Fix Flux tuning issue by @mengniwang95 in #936
• gguf support for inclusionAI/Ling-flash-2.0 by @n1ck-guo in #940
• remove low_cpu_mem by @n1ck-guo in #934
• Add compatibility test by @XuehaoSun in #918
• Add commit hash to version by @XuehaoSun in #941
• gguf weight type align with original, output.weight, token_embed by @n1ck-guo in #900
• support attention mask in user's dataset by @wenhuach21 in #930
• Add diffusion README by @mengniwang95 in #923
• update readme by @wenhuach21 in #949
• refactor utils file by @n1ck-guo in #943
• update readme for sglang support by @WeiweiZhang1 in #953
• update gguf and support for CompressedLinear by @n1ck-guo in #950
• Reduce AutoSchem VRAM usage by up to 10X by @wenhuach21 in #944
• add self attribution and fix avg_bits error by @xin3he in #956
• add logo by @wenhuach21 in #960
• refine AutoScheme readme/code by @wenhuach21 in #958
• update readme by @wenhuach21 in #962
• fix critic disable_opt_rtn regression by @wenhuach21 in #963
• [1/N] Initial vllm-ext evaluation support (MXFP4 MOE) by @yiliu30 in #935
• fix bug of imatrix contains 0 by @n1ck-guo in #955
• fix rtn bug by @mengniwang95 in #966
• enhance flux doc by @mengniwang95 in #967
• clean code by @wenhuach21 in #968
• support for model scope by @n1ck-guo in #957
• merge main branch to alg_ext by @wenhuach21 in #970
• fix cuda CI backend issue, fixtypo by @WeiweiZhang1 in #974
• disable compile packing by default by @yiliu30 in #975
• enhance auto device map and support XPU by @xin3he in #961
• refine readme by @wenhuach21 in #978
• cli support for positional arguments model by @n1ck-guo in #979
• update bits in UT by @xin3he in #986
• fix guff scheme and device_map bug by @n1ck-guo in #969
• add support for Magistral-Small by @n1ck-guo in #980
• support model_dtype and fix bug of scheme contains quotes, mllm eval by @n1ck-guo in #985
• fix bug of cannot create adam compressor by @n1ck-guo in #992
• [CI] Update python to 3.12 and torch to 2.8.0 by @XuehaoSun in #741
• fix lm head bug and rm clear_mem_reach_threhold by @wenhuach21 in #997
• Reduce peak gpu memory usage and support moe estimation by @xin3he in #981
• fix cuda ut bug by @n1ck-guo in #999
• fix mllm device_map ut by @Kaihui-intel in #1000
• refine md tables by @WeiweiZhang1 in #994
• Refine exllamav2 ut by @WeiweiZhang1 in #1001
• Support for immediate saving to reduce ram usage by @Kaihui-intel in #965
• Fix diffusion multi-device ut issue by @mengniwang95 in #1002
• fix multiple devices map issue in calibration by @wenhuach21 in #1003
• Fix non auto device map by @WeiweiZhang1 in #1005
• fix multiple devices issue in Compressor and AutoScheme by @wenhuach21 in #1007
• fix cuda low_cpu_mem_usage ut by @Kaihui-intel in #1010
• Fix param missing bug by @mengniwang95 in #1008
• add device list to clear memory by @wenhuach21 in #1009
• Minor refactor for LLMC by @yiliu30 in #993
• fix one clear memory issue by @wenhuach21 in #1011
• add ut for gguf alg_ext and update so file by @n1ck-guo in #1012
• fix multi cuda ut bug by @n1ck-guo in #1014
• Including auto_scheme.default_alg into pypi by @chensuyue in #1018
• add num_device check for set_auto_device_map_for_block_with_tuning by @xin3he in #1021
• dispatch model with real max memory by @xin3he in #1022
• fix cuda ut by @n1ck-guo in #1020
• disable itrex format first by @WeiweiZhang1 in #998
• fix bug of lm_head and dispatch model,gguf eval by @n1ck-guo in #1025
• Fix the missing temporary name by @yiliu30 in #1029
• Reduce mem usage of GPT-OSS by @yiliu30 in #1013
• update gguf alg ext by @n1ck-guo in #1026
• optimize vram for gguf and add momentum by @wenhuach21 in #1031
• fix incorrect model name in readme by @wenhuach21 in #1035
• Bump into v0.9.0 by @XuehaoSun in #1024

Full Changelog: v0.8.0...v0.9.0

v0.8.0

Highlights

• merge all api(MLLM, Adam) into AutoRound by @n1ck-guo in #791
• MXFP4 and MXFP8 loading support by @yiliu30 in #832
• Support Flux quantization by @mengniwang95 in #850

What's Changed

• fix cuda ut bug of use_deterministic_algorithms by @n1ck-guo in #805
• remove torch compile in nv quant by @wenhuach21 in #807
• Support loading for static quant weight fp8 act fp8 by @yiliu30 in #730
• fix bug of q_layer_inputs by @n1ck-guo in #811
• fix gptqmodel inference issue by @wenhuach21 in #813
• Bump version to v0.7.0 by @XuehaoSun in #814
• fix nsamples in get_dataloader by @wenhuach21 in #804
• Refine logger and add envs by @yiliu30 in #817
• Fix llm-compressor export by @Kaihui-intel in #820
• enhance auto-round eval with vllm backend by @xin3he in #815
• rm triton from requirements and correct the supported python version to 3.10(+) by @wenhuach21 in #824
• move environment variable setting into eval function by @xin3he in #829
• bump version to 0.8.0.dev by @XuehaoSun in #830
• [STEP 1] merge all api(MLLM, Adam) into AutoRound by @n1ck-guo in #791
• add support for scheme FP8_STATIC to export llm_compressor format by @n1ck-guo in #816
• fix format checking bug by @WeiweiZhang1 in #836
• MXFP4 and MXFP8 loading support by @yiliu30 in #832
• hpu build with auto_round package name by @chensuyue in #838
• fix hpu detect issue by @xin3he in #823
• fix severe vram leak regression in auto-round format packing by @wenhuach21 in #842
• fix tp device issue caused by device_map by @xin3he in #833
• fix log error by @n1ck-guo in #843
• [High Risk]Refine inference code by @wenhuach21 in #840
• fix gguf fp8 input model and vram issue by @wenhuach21 in #844
• NVFP4 Loading support by @yiliu30 in #839
• fix extra config by @n1ck-guo in #847
• change the method of detecting linear by @n1ck-guo in #849
• fix device_map setting by @Kaihui-intel in #854
• Add typo checker by @XuehaoSun in #846
• fix parse layer config bug by @wenhuach21 in #856
• Refine BackendInfo to include act fields by @yiliu30 in #848
• fix bug of data_type fp8_sym by @n1ck-guo in #855
• fix save_quantied format cheaker by @WeiweiZhang1 in #857
• fix bug of get_layer_names_in_block by @wenhuach21 in #861
• raise vlm loading error by @wenhuach21 in #863
• fix FP8 model as input and backend issue by @wenhuach21 in #864
• fix seqlen bug and calib slow of mllm tuning by @n1ck-guo in #871
• fix device bug by @xin3he in #873
• fix vllm backend evaluation by @xin3he in #872
• Optimize CPU unit test workflow by @XuehaoSun in #881
• Fix Cuda CI failures due to Transformers and AWQ incompatibility by @WeiweiZhang1 in #882
• Support Flux quantization by @mengniwang95 in #850
• fp8 exporting bugfix by @WeiweiZhang1 in #874
• lm_eval stop try except and add back missing arguments by @xin3he in #884
• Fix act calibration bug by @mengniwang95 in #880
• restrict accelerate version by @wenhuach21 in #885
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in #868
• update require accelerate version by @n1ck-guo in #888

Full Changelog: v0.7.1...v0.8.0

v0.7.1 patch release

fix severe vram leak regression in auto-round format packing @ #842

v0.7.0

🚀 Highlights

• Enhanced NVFP4 algorithm and added support to export MXFP4/NVFP4 to the llm-compressor format
  by @WeiweiZhang1 and @wenhuach21

• Improved W2A16 quantization algorithm
  by @wenhuach21

• Introduced the scheme interface for easier configuration of quantization settings
  by @wenhuach21

• Added support for using FP8 models as input and str name as model input in API
  by @wenhuach21 and @n1ck-guo

• Unified device and device_map arguments and introduced device_map="auto"
  to simplify quantization of extremely large models
  by @Kaihui-intel

What's Changed

• fix ut import issue by @WeiweiZhang1 in #686
• support to export static afp8 model by @n1ck-guo in #662
• Add ruff and isort by @XuehaoSun in #578
• Improved log message for unsupported dataset by @wenhuach21 in #688
• support rceil for mxfp by @wenhuach21 in #660
• Add black and blacken-docs in pre-commit by @XuehaoSun in #692
• support static global scale for nvfp4 and update readme by @wenhuach21 in #691
• Update readme by @wenhuach21 in #695
• Add script for cuda unit test by @XuehaoSun in #567
• support to save image_processor by @n1ck-guo in #694
• support for static activation quantization calibration with group_size by @n1ck-guo in #693
• fix xpu oom checker by @n1ck-guo in #705
• FIXBUG: CPU Offloading for Cache Blocks in Low-Memory GPU Systems or Single GPU on ROCM Configs by @JartX in #703
• fix bug of zero accuracy for mx-fp by @n1ck-guo in #709
• catch oom error and move to cpu directly by @n1ck-guo in #708
• code optimization of vlm by @n1ck-guo in #704
• fix critic bug of gguf tuning by @wenhuach21 in #710
• support fp8 model and str as input in llm quantization by @wenhuach21 in #699
• change act_scale to input_scale for fp8 export by @n1ck-guo in #711
• simply CpuInfo class by @wenhuach21 in #715
• Update step_by_step.md by @wenhuach21 in #717
• fix bug of activation quant when act_max is None by @n1ck-guo in #718
• Bump transformers in /test/test_cuda by @dependabot[bot] in #719
• Freeze torchvision version in CI by @XuehaoSun in #720
• update autoround mllm and support Mistral 3.2 series by @n1ck-guo in #713
• Fix hpu CI by @XuehaoSun in #723
• fix fp8 model input issue by @wenhuach21 in #724
• update gguf convert.py and support for gpt-oss by @n1ck-guo in #721
• new cast_to_nvfp4 with high performance by @xin3he in #727
• make the tuning deterministic and move infrequently used arguments to kwargs by @wenhuach21 in #726
• add original convert file and support for the newest llama.cpp by @n1ck-guo in #729
• fix bug for exporting afp8 fake format by @n1ck-guo in #731
• Fix torch_zp infer bug & API disable_deterministic_algorithms bug by @WeiweiZhang1 in #733
• fix gguf mistral_common import by @n1ck-guo in #736
• Enable mxfp exporting by @WeiweiZhang1 in #649
• support for glm4.5 gguf by @n1ck-guo in #735
• support auto-round-mllm command by @n1ck-guo in #742
• Optimize pack zeros for int sym by @WeiweiZhang1 in #743
• fix UT check for int zp by @WeiweiZhang1 in #745
• support llama4 quant by @mengniwang95 in #744
• fix bug of loading fp8 model by @n1ck-guo in #747
• improved algorithm for int2 by @wenhuach21 in #748
• Add Static FP8 KV Support by @yiliu30 in #737
• refine code by @wenhuach21 in #749
• mllm supports loading fp8 model and fix bug of loading fp8 model by @n1ck-guo in #750
• support deepspeed LinearLayer and LinearAllreduce by @xin3he in #698
• fix alg_ext moe and model str input bug by @wenhuach21 in #751
• api support for fp8 model and mllm api support load from str by @n1ck-guo in #752
• fix some torch compile warnings by @wenhuach21 in #755
• Speedup FP4 packing by @yiliu30 in #760
• fix_script_fp_layer_config_for_bits_checking by @WeiweiZhang1 in #756
• support quant lm_head for rtn w8afp8 static quant by @n1ck-guo in #754
• Revert "Speedup FP4 packing" by @yiliu30 in #763
• refine code and fix activation quantization eval regression by @wenhuach21 in #762
• fix gguf ut bug by @n1ck-guo in #767
• fix gguf bug by int zp by @n1ck-guo in #771
• Keep the model’s buffer dtype unchanged in most cases by @wenhuach21 in #770
• fix set_layer_config bug by @wenhuach21 in #768
• fix bug of auto_round exporting by @n1ck-guo in #772
• gguf format supports for fp8 model by @n1ck-guo in #778
• [API CHANGE] Stage 1 add quant scheme and consolidate device and device_map by @wenhuach21 in #774
• Speedup FP4 packing by @yiliu30 in #766
• hot fix for nvfp4 scheme by @wenhuach21 in #784
• fix alg_ext regression and support mxfp4 in it with slight improvement by @wenhuach21 in #785
• refine nvfp code, typofix by @WeiweiZhang1 in #777
• mxfp/nvfp/fp8 support torch compile in tuning by @wenhuach21 in #789
• refine nvfp4 algorithm by @wenhuach21 in #790
• add limit arg for eval by @n1ck-guo in #764
• torch backend bugfix and speedup ut by @WeiweiZhang1 in #793
• Support auto device mapping by @Kaihui-intel in #781
• fix bug and add nvfp in alg-ext with slight improvement by @wenhuach21 in #794
• rename llmcompressor to llm_compressor for align with other formats by @WeiweiZhang1 in #780
• align formats packing device to API by @WeiweiZhang1 in #795
• add fp8 export format check by @n1ck-guo in #779
• fix several regressions including lm-head quantization, 3bit asym torch backend,etc by @wenhuach21 in #796
• refine readme by @wenhuach21 in #798
• fix typo in readme by @wenhuach21 in #799
• fix several cuda ut bug by @n1ck-guo in #797
• enable model python files saving by @WeiweiZhang1 in #802
• AutoRoundMLLM supports scheme and fix device_map=dict regression by @n1ck-guo in #801
• improve the robustness of scheme by @wenhuach21 in #803
• fix mxfp exporting by @WeiweiZhang1 in #806

New Contributors

• @JartX made their first contribution in #703
• @mengniwang95 made their first contribution in #744

Full Changelog: v0.6.0...v0.7.0

v0.6.0

Highlights

• provide experimental support for gguf q*_k format and customized mixed bits setting
• support xpu in triton backend by @wenhuach21 in #563
• add torch backend by @WeiweiZhang1 in #555
• provide initial support of llmcompressor format, only INT8 W8A8 dynamic quantization is supported by @xin3he in #646

What's Changed

• bump version into v0.5.1 by @XuehaoSun in #540
• Freeze pytorch & ipex version in CI by @XuehaoSun in #541
• fix_quantization_config_for_inference by @WeiweiZhang1 in #542
• [critic bug]remove redundant round in dq simulation by @wenhuach21 in #543
• update readme by @wenhuach21 in #550
• add recipes for qwen3 8b and 14b by @n1ck-guo in #552
• itrex requires torch<2.7 by @XuehaoSun in #548
• [GGUF STEP4] fix search bug and improve packing & eval speed by @n1ck-guo in #545
• refine xpu requirement/config json and fix several issues by @wenhuach21 in #558
• add UE5M3 simulation by @wenhuach21 in #562
• support xpu in triton backend by @wenhuach21 in #563
• fix typo in backend by @wenhuach21 in #564
• update habana docker to 1.21.0 by @XuehaoSun in #566
• Support for more gguf format and float zp for Q*_1 by @n1ck-guo in #560
• update readme by @wenhuach21 in #569
• update readme by @wenhuach21 in #571
• support for llava-based hf model by @n1ck-guo in #568
• add gguf accuracy data by @wenhuach21 in #574
• add sym & asym gguf quant for gguf baseline (iter==0) by @n1ck-guo in #573
• modify default asym 4bits auto-round format to awq, fix save folder typo for mllm by @WeiweiZhang1 in #575
• improve the robustness of parsing vlm config by @wenhuach21 in #577
• switch to transformers API in cpu ut by @wenhuach21 in #580
• add torch backend by @WeiweiZhang1 in #555
• fix awq exporting at group_size=-1 by @wenhuach21 in #579
• refact cuda ut to facilitate automation by @n1ck-guo in #559
• fix tensor shape mismatch error for API usage by @WeiweiZhang1 in #582
• fix device bug at calibration by @wenhuach21 in #587
• Update gguf_accuracy (q3_ks) by @SinpackKonmakan in #590
• add recipes for deepseek-r1-0528 by @n1ck-guo in #588
• correct errors of deepseek-r1-0528 recipes by @n1ck-guo in #591
• fix cuda ut by @wenhuach21 in #592
• Bump protobuf from 3.20.1 to 3.20.2 in /test/test_cuda by @dependabot[bot] in #585
• rm unnecessary forward to improve speed by @wenhuach21 in #593
• update readme by @wenhuach21 in #597
• fix q2k bug by @n1ck-guo in #599
• support for q4_k_m by @n1ck-guo in #596
• fix vlm uttest path error by @WeiweiZhang1 in #601
• fix lots of gguf critic bugs and support imatrix in rtn mode by @wenhuach21 in #595
• fix gguf bug by @wenhuach21 in #610
• mv some checkers by @wenhuach21 in #611
• fix gguf packing bug and moe regression by @wenhuach21 in #614
• support customized mixed bits for gguf by @wenhuach21 in #615
• fix double quant sym bug by @wenhuach21 in #616
• FP8 WOQ export by @wenhuach21 in #617
• fix bug of q5_k_s w/ imatrix by @n1ck-guo in #620
• add auto-round related vllm and transformers UT by @WeiweiZhang1 in #613
• refine_doc_0624 by @WeiweiZhang1 in #619
• fix not using imatrix for gguf at rtn mode by @wenhuach21 in #623
• fix vlm hf config loading issue by @WeiweiZhang1 in #624
• refine gguf rtn algorithm and fix bugs by @wenhuach21 in #630
• fix gguf bug of moe models and lmhead/embedding bits setting regression by @n1ck-guo in #628
• [BUG FIX] fix bug of deepseek gguf:q*k by @n1ck-guo in #637
• support packing immediately for gguf to reduce ram usage by @wenhuach21 in #638
• support llmcompressor format by @xin3he in #646
• fix norm_bias_tuning by @wenhuach21 in #639
• [W4A8]Fix Packing by @yiliu30 in #648
• Integrate RTN quantization into GGUF packing to enhance robustness by @n1ck-guo in #644
• Remove vlm cuda UT dependencies version restrictions by @XuehaoSun in #651
• speedup mxfp tuning and fix nvfp bug by @wenhuach21 in #647
• support two more calib datasets and fix embedding layer bug by @wenhuach21 in #653
• fix some issues by @wenhuach21 in #655
• fix bug of q4_0 and q5_0 at iters==0 by @n1ck-guo in #658
• support vlm models for gguf format by @n1ck-guo in #654
• fix bug of block-wise quant imatrix by @n1ck-guo in #663
• fix gguf block-wise issue by @wenhuach21 in #664
• fix bugs of export deepseek gguf format when iters=0 and q3k accuracy by @n1ck-guo in #665
• handle zeros in imatrix by @wenhuach21 in #667
• fix ut issue by @WeiweiZhang1 in #668
• fix cuda hanging issue during packing by @WeiweiZhang1 in #669
• support to use lm_eval for vlm by @n1ck-guo in #670
• add trust remote code to gguf format load tokenizer by @n1ck-guo in #675
• fix 3bits asym accuracy and calib dataset issues by @WeiweiZhang1 in #674
• restrict accelerate version to reduce ram usage by @wenhuach21 in #673
• rm low_cpu when loading the model by @wenhuach21 in #676
• rm_old_vlm_cuda_ut by @WeiweiZhang1 in #678
• update gguf convert file and fix bug of permute bug by @n1ck-guo in #679
• fix gguf regression for large models by @wenhuach21 in #680
• fix gemma vlm gguf regression by @wenhuach21 in #685

New Contributors

• @SinpackKonmakan made their first contribution in #590
• @xin3he made their first contribution in #646

Full Changelog: v0.5.1...v0.6.0

v0.5.1:bug fix release

What's Changed

• bump version into v0.5.0 by @XuehaoSun in #538
• fix triton multiple gpus and some other issues by @wenhuach21 in #539

Full Changelog: v0.5.0...v0.5.1

v0.5.0

Highlights

• refine autoround format inference, support 2,3,4,8 bits and marlin kernel and fix several bugs in auto-round format
• support xpu in tuning and inference by @wenhuach21 in #481
• support for more vlms by @n1ck-guo in #390
• change quantization method name and made several refinements by @wenhuach21 in #500
• support rtn via iters==0 by @wenhuach21 in #510
• fix bug of mix calib dataset by @n1ck-guo in #492

What's Changed

• support xpu in tuning and inference by @wenhuach21 in #481
• add light ut, fixtypos by @WeiweiZhang1 in #483
• bump into v0.4.7 by @XuehaoSun in #487
• fix dataset combine bug by @wenhuach21 in #489
• fix llama 8b time cost by @WeiweiZhang1 in #490
• update 2bits acc results by @WeiweiZhang1 in #491
• fix bug of mix calib dataset by @n1ck-guo in #492
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci in #494
• [GGUF support step3]patch for double quant by @n1ck-guo in #473
• refine inference backend/code step 1 by @wenhuach21 in #486
• refine inference step 2 by @wenhuach21 in #498
• change quantization method name and made several refinements by @wenhuach21 in #500
• fix bug of awq/gptq modules_to_not_convert by @n1ck-guo in #501
• use --tasks to control evaluation enabling by @wenhuach21 in #505
• fix gguf eval regression bug by @n1ck-guo in #506
• change to new api in readme by @wenhuach21 in #507
• fix setup issue on cuda machine by @wenhuach21 in #511
• support rtn via iters==0 by @wenhuach21 in #510
• fix critical bug of get_multimodal_block_names by @n1ck-guo in #509
• Update requirements-lib.txt by @yiliu30 in #513
• add group_size divisible check in backend by @wenhuach21 in #512
• support for more vlms by @n1ck-guo in #390
• move gguf-dq test to cuda by @n1ck-guo in #520
• fix bs!=1 for gemma and MiniMax-Text-01 by @wenhuach21 in #515
• add regex support in layer_config setting by @wenhuach21 in #519
• patch for vlm by @n1ck-guo in #518
• rename backend to packing_format in config.json by @wenhuach21 in #521
• fix example's model_dtype by @WeiweiZhang1 in #523
• rm fp16 export in autoround format by @wenhuach21 in #525
• update convert_hf_to_gguf to support more models by @n1ck-guo in #524
• fix light config by @WeiweiZhang1 in #526
• fix typos, add model card link for VLMs by @WeiweiZhang1 in #527
• add backend readme by @wenhuach21 in #528
• update mllm readme by @WeiweiZhang1 in #530
• fix bug of cuda ut by @n1ck-guo in #532
• fix inference issue by @wenhuach21 in #529
• update readme by @wenhuach21 in #531
• refine readme by @WeiweiZhang1 in #536
• fix cuda ut by @n1ck-guo in #537

Full Changelog: v0.4.7...v0.5.0

v0.4.7

Highlights

Support W4AFP8 for HPU. Please refer to Intel Neural Compressor for guidance on running these models. by @yiliu30 in #467

Support packing immediately in new quantization api to save ram usage by @wenhuach21 in #466

20x for awq and 4x for gptq packing speedup on cuda by @wenhuach21 in #459

Support auto-round-light to speed up the tuning process @WeiweiZhang1 in #454

Fix critic bug of mxfp4 in tuningby @wenhuach21 in #451

What's Changed

• step-1 support naive double quant in tuning by @wenhuach21 in #442
• fix critic bug of mxfp4 by @wenhuach21 in #451
• update readme by @wenhuach21 in #455
• update eval by @n1ck-guo in #450
• awq exporting bugfix by @WeiweiZhang1 in #456
• Support force loading into autoround Format by @WeiweiZhang1 in #453
• 20x for awq and 4x for gptq packing speedup by @wenhuach21 in #459
• fixl eval bug by @n1ck-guo in #461
• [STEP-1]W4Afp8 export by @wenhuach21 in #378
• [HPU] Update W4A8 for HPU by @yiliu30 in #467
• support for gemma3 by @n1ck-guo in #468
• upload_auto-round-light results by @WeiweiZhang1 in #454
• GGUF support step2: add naive Q2_KS and Q4_KS by @n1ck-guo in #448
• fix incorrect recipe data by @WeiweiZhang1 in #471
• support for mistral3 by @n1ck-guo in #472
• support to export gemma3 gguf format by @n1ck-guo in #470
• Increase unit test timeout from 120 to 240 minutes by @XuehaoSun in #474
• support packing immediately in new quantization api to save ram usage by @wenhuach21 in #466
• rm redundant line break by @WeiweiZhang1 in #475
• Temporarily close qxk api for new release by @n1ck-guo in #478
• add restrict for exporting act-quant models by @n1ck-guo in #480

Full Changelog: v0.4.6...v0.4.7