Jun Jiang

I will rewrite MoyaX to full compatibility with Swift 3.0, maybe a week later, for now I'm organizing RubyConf China, haven't much time to do this job.

Pagy 8 support

4

This PR mixed a few of things: - Loosen requirement to support Pagy 8 - Fix typo - Slightly restructure file organization - bump version - Add Ruby 3.3 to...

According to https://developers.deepl.com/docs/api-reference/translate/openapi-spec-for-text-translation `source_lang` is optional, and DeepL will auto-detect language if it's not given. Note: I don't have Unity or C# development environment, so I can't compile it, but...

Jetson (aarch64) support

7

I refactored `setup.py` to make it work on my Jetson AGX Orin, I think it also helps for future ARM + GPU platforms I don't want to make it complex...

Any plan to support WASM target?

2

Do you have a plan to support the WASM target? The use case is similar to Smoldot.

Switch back to NVIDIA/cutlass, and upgrade to v4.2.1

13

The issue has been fixed https://github.com/NVIDIA/cutlass/blob/v4.2.1/tools/util/include/cutlass/util/mixed_dtype_utils.hpp#L180-L184 So I switch back to the official repo and upgrade to v4.2.1, which is important for ARM Blackwell

It seems xformers/csrc/sparse24/sparse24_gemm_sm90.cu not follow TMA alignment

1

# 🐛 Bug I'm trying to upgrade cuTLASS to the latest 4.2 for xformers, when compiling, I meet an error ``` FAILED: [code=255] /root/tmp/xformers/build/temp.linux-x86_64-cpython-312/xformers/csrc/sparse24/sparse24_gemm_sm90.o /usr/local/cuda/bin/nvcc --generate-dependencies-with-compile --dependency-output /root/tmp/xformers/build/temp.linux-x86_64-cpython-312/xformers/csrc/sparse24/sparse24_gemm_sm90.o.d -I/root/tmp/xformers/xformers/csrc -I/root/tmp/xformers/third_party/sputnik...