#moearchitecture
DeepEP: New high-performance MoE communication library with advanced GPU kernels and RDMA support
https://github.com/deepseek-ai/DeepEP
#gpucomputing #aiinfrastructure #networking #moearchitecture #performanceoptimization
February 25, 2025 at 5:34 AM
PyTorch profiling data reveals optimization strategies for MoE layers in training and inference
https://github.com/deepseek-ai/profile-data
#performanceprofiling #deeplearning #moearchitecture #pytorch #parallelcomputing
February 27, 2025 at 7:21 PM