"examples/offline_inference/data_parallel.py" did not exist on "7f6bae561c210da06af5d40e8861b0d2ddfe339c"
-
zhuwenwen authored
set default_max_num_batched_tokens = 10240 update qwen3_moe of layernorm off lightop of moe_fused_gate
e7f2785f
set default_max_num_batched_tokens = 10240 update qwen3_moe of layernorm off lightop of moe_fused_gate