5 Pull requests merged by 4 people

• Port metadata from the linear node onto the reference custom op for int4

  #2860 merged Aug 29, 2025

• Add tracking for new tensors, AQT and layouts

  #2895 merged Aug 29, 2025

• safetensors support

  #2881 merged Aug 29, 2025

• [mxfp8 moe training] add per group blocked scale kernels

  #2886 merged Aug 28, 2025

• Rename to_float8 to from_hp

  #2893 merged Aug 28, 2025

7 Pull requests opened by 6 people

• hf integration doc page

  #2899 opened Aug 28, 2025

• [mxfp8 moe training] integrate triton kernels for converting scales to blocked format

  #2902 opened Aug 28, 2025

• [moe training] update tests + benchmarks with conditional runs based on SM arch; make test cases more comprehensive and consistent

  #2905 opened Aug 28, 2025

• Add AWQ-INT4 option to release script

  #2906 opened Aug 29, 2025

• Fix FX Graph Cache issue in register_da8w4_concat_linear_cpu_pass

  #2907 opened Aug 29, 2025

• torchao init: do not load .so files for known incompatible torch version

  #2908 opened Aug 29, 2025

• Remove unused cpp variable, breaking style checks

  #2909 opened Aug 29, 2025

1 Issue closed by 1 person

• Using triton_op + wrap_triton introduces kernel performance regression

  #2898 closed Aug 28, 2025

3 Issues opened by 3 people

• float8 rowwise scaled grouped mm doesn't support B200

  #2904 opened Aug 28, 2025

• Loading fp8-int4 model got unexpected keyword argument 'requires_grad'

  #2903 opened Aug 28, 2025

• Aborted (core dumped) when importing v0.13.0 RC

  #2901 opened Aug 28, 2025

16 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• Make SmoothQuant more General

  #2728 commented on Aug 29, 2025 • 26 new comments

• refactor common used toy model

  #2729 commented on Aug 29, 2025 • 21 new comments

• Add Int4PlainInt32 Tensor

  #2845 commented on Aug 29, 2025 • 19 new comments

• Fix Float8Tensor quantize op kernrel preference dispatch

  #2883 commented on Aug 29, 2025 • 5 new comments

• [mxfp8 moe training] add triton kernel for blocked swizzled 3d weight scales

  #2894 commented on Aug 28, 2025 • 4 new comments

• Move CPU kernels out of experimental

  #2868 commented on Aug 29, 2025 • 3 new comments

• CUDA OOM When Running AWQ int4 Quantized llama3.1-8b at Batch Size 1

  #2867 commented on Aug 28, 2025 • 0 new comments

• Benchmark AWQ and SmoothQuant within vLLM ecosystem

  #2815 commented on Aug 28, 2025 • 0 new comments

• CI ROCM tests failing with "HW exception - GPU hang"

  #2890 commented on Aug 28, 2025 • 0 new comments

• Migrating from AffineQuantizedTensor + Layouts to new structure of tensor subclasses

  #2752 commented on Aug 29, 2025 • 0 new comments

• [CPU][FP8][Inductor] How to support fp8 quant for inductor on CPU

  #2896 commented on Aug 29, 2025 • 0 new comments

• [CPU] Add support for dynamic float8 act float8 weight on CPU

  #2505 commented on Aug 29, 2025 • 0 new comments

• Replace `torch.norm` with `torch.linalg.vector_norm`

  #2660 commented on Aug 29, 2025 • 0 new comments

• Add Int4TilePackedTo4dTensor

  #2791 commented on Aug 29, 2025 • 0 new comments

• Float8Tensor per row quantization pass bias to fbgemm kernel

  #2884 commented on Aug 29, 2025 • 0 new comments

• [mxfp8 moe training] use dim1 cast cuda kernel in bwd

  #2897 commented on Aug 28, 2025 • 0 new comments