laekov/fastmoe Issues
MoE L2 norm reduce in Megatron
Closed 3DDP error
ClosedA bug in switch_gate
Updated 6About switch_gate
Updated 1multi-node problem
Updated 1你好,我想请问下在fastmoe中如何定义 dp+mp下的moe
Closed 6Example to run Megatron
Updated 3请问fastmoe能被集成到VLLM里吗
Updated 4跑FMOE的时候提示cudaErrorInvalidDevice
Closed 6fastmoe支持微调吗
Closedprep_text8.py没有该脚本
Closed 1我们有线上沟通的群吗
Updated 1pytest error
Updated 3setup.py error!
Closed 4how to use balance loss?
Updated 1Distributed Training is failing
Closed 9MoE DDP + Expert Parallelism
Closed 6More GPU number than expert number
Closed 5CUBLAS_STATUS_ARCH_MISMATCH
Closed 2About balance loss
Closed 3