zyksir 8e3797be1c support 1 shot allreduce in 1-node and 2-node using mscclpp (#6277)

2025-06-04 22:11:24 -07:00

..

all_reduce

support 1 shot allreduce in 1-node and 2-node using mscclpp (#6277)

2025-06-04 22:11:24 -07:00

decoding_attention_triton

…

deepep

…

deepseek

…

fused_moe_triton

Set num_fused_shared_experts as num_shared_experts when shared_experts fusion is not disabled (#6736)

2025-06-04 15:53:22 -07:00

minmax-text-01-lightning_attention

…

quantization

…

rmsnorm

…

scheduler_batch

[test] add ut and bm for get_last_loc (#6746)

2025-05-29 11:47:21 -07:00