config: derive the runner's DCP topology from its ParallelState (#34133)
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
This commit is contained in:
co-authored by
Claude Opus 5
parent
22e003580b
commit
ce1b9f88b6
@@ -322,13 +322,14 @@ def load_model(server_args, port_args, gpu_id, tp_rank):
|
||||
attn_tp_size=attn_tp_size,
|
||||
attn_cp_rank=0,
|
||||
attn_cp_size=server_args.attn_cp_size,
|
||||
attn_dcp_rank=tp_rank % server_args.dcp_size,
|
||||
attn_dcp_size=server_args.dcp_size,
|
||||
attn_dp_rank=attn_dp_rank,
|
||||
attn_dp_size=attn_dp_size,
|
||||
moe_ep_rank=moe_ep_rank,
|
||||
moe_ep_size=server_args.ep_size,
|
||||
moe_dp_rank=None,
|
||||
moe_dp_size=server_args.moe_dp_size,
|
||||
dcp_size=server_args.dcp_size,
|
||||
gpu_id=gpu_id,
|
||||
)
|
||||
runner_kwargs = dict(
|
||||
|
||||
Reference in New Issue
Block a user