alankelly

Results 49 comments of alankelly

QB4W SSE2/SSE41 GEMM Kernels

Please resolve conflicts and I will import

Enable WASM build on GitHub Actions

Hi, a pull request for this would be most welcome! Probably involves editing: https://github.com/google/XNNPACK/blob/master/.github/workflows/build.yml Thanks!

Fix logging errors

Thanks for this, sorry it slipped through the cracks. Importing now

[TFLite] Fix Android build with CMake

@gbaned I no longer work on TFLite. Please assign to someone else.

Enable wasmusdot qd8_f32_qc8w gemm/igemm in gemm-config

Can you please delete the amalgam? They are now automatically generated.

LTO type mismatch on `xnn_table_exp2minus_k_over_*` kernels

Could you please open a PR with a fix?

how to build with enabled sme arm kleidai

You need to pass these flags to blaze as it is not enabled by default: --define xnn_enable_arm_sme=true --define xnn_enable_arm_sme2=true We have tested this on iOS. You must create an XCode...

converting NER model with transformer into tflite

Have you tried resizing the input tensor? https://www.tensorflow.org/api_docs/python/tf/lite/Interpreter#resize_tensor_input

Plans for ARM SME support?

F32 SME support is added in https://github.com/google/XNNPACK/pull/7260 Support for other types will come in the next weeks/months