MtFitzRoy

Results 5 comments of MtFitzRoy

When can we support qwen3 bf16 with deepep?

@yizhang2077 Seems DeepEP already support bf16 dispatch https://github.com/deepseek-ai/DeepEP?tab=readme-ov-file#roadmap

@yizhang2077 For normal mode dispatch, I think sglang code already support bf16.

@zhyncs Hi, Do we multi MTP heads now? Is there an example?

Is there a plan to support TP + SP attention? The paper says "The attention part employs 4-way Tensor Parallelism (TP4) with Sequence Parallelism (SP)"