yangulei opened a new pull request, #11111: URL: https://github.com/apache/tvm/pull/11111
Enable [bfloat16](https://discuss.tvm.apache.org/t/rfc-add-bfloat16-data-type/6778) in [DNNL BYOC](https://discuss.tvm.apache.org/t/rfc-byoc-intel-r-onednn-integration/11582) following the path: > [float32 graph] --> \<[AMP](https://discuss.tvm.apache.org/t/rfc-relay-fp32-fp16-model-support/9994)\> --> [bfloat16 graph] --> \<BYOC\> --> [TVM + oneDNN module] Main work include: - Enable more data types in DNNL json runtime (only bfloat16 has been tested so far). - Consider dtype while querying optimal DNNL layout. - Add tests for bf16 DNNL BYOC. With those improvements, a float32 graph could be converted to bfloat16 through AMP, and then be lowered by native codegen or consumed by oneDNN and finally inference in bfloat16 mode now. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
