Xin He

Signed-off-by: Xin He ## Type of Change add docstring for algorithms.

support habana FP8 per channel quantization

2

## Type of Change feature ## Description - [x] support per channel quantization for higher accuracy - [x] add observer registry for easy extension - [x] dump scale_inv from observer...

[RFC] Porting INC SmoothQuant recipes to IPEX autotune API

2

https://github.com/intel-innersource/frameworks.ai.pytorch.ipex-cpu/issues/2404

fix performance issue in mistral

3

# What does this PR do? Fix performance issue in mistral Without this fix, the first token generation will take more time and the performance is bad. ## Before submitting...

Hi @casper-hansen, For general GEMM quant type, I observe that the `qweight `shape of AutoAWQ and AutoGPTQ is different due to different pack dimension. I'm confused about why we introduce...

Bad result when running AWQ without GPU

4

Hi, folks, I met some weird issue when reproducing the results shown in paper. I can get results below with GPU visible, but cannot reproduce it with only CPU. I...

implement `incbench` command for ease-of-use benchmark

3

## Type of Change feature ## Description - [x] implement `incbench` command as entrypoint for ease-of-use benchmark - [x] automatically check numa/socket info and dump it with table for ease-of-understand...

fix bf16 symbolic_trace bug

1

## Type of Change bug fix ## Description fix bf16 symbolic_trace bug, 1. cause abnormal recursive calling. 2. missing necessary attributes By moving BF16 fallback ahead of quantization and removing...

lm_head is not converted to QuantLinear with MXFP4/8

14

lm_head quantization still have some issues. - need deepcopy if tied_word_embedding = True - export is not applied for lm_head Shall we warn user that lm_head is not supported? @WeiweiZhang1...