DickJC123 opened a new pull request #7505: Changed FullyConnected to use new linalg gemm, plus TensorCore if fp16 I/O. URL: https://github.com/apache/incubator-mxnet/pull/7505 GEMMs within the FullyConnected operator switched from using mshadow::dot() to the new linalg_gemm(). After a trial-run, this can be the model for removing all uses of dot() within MXNet. Added a specialization linalg_gemm<gpu, half_t> that includes use of TensorCore algos by default. Users can disable TensorCore on Volta by setting the environment variable MXNET_CUDA_ALLOW_TENSOR_CORE=0. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org
With regards, Apache Git Services