Unverified Commit 546914fa authored by XSongQ's avatar XSongQ Committed by GitHub
Browse files

[Fix] Fix the bug of the calculation of base_gpu_id (dp offset) in...

[Fix] Fix the bug of the calculation of base_gpu_id (dp offset) in data_parallel_controller.py (#10741)
parent 4726c919
......@@ -209,7 +209,9 @@ class DataParallelController:
args=(server_args, tmp_port_args, base_gpu_id, dp_rank, ready_event),
)
threads.append(thread)
base_gpu_id += server_args.tp_size * server_args.gpu_id_step
base_gpu_id += (
server_args.tp_size * server_args.pp_size * server_args.gpu_id_step
)
# Free all sockets before starting the threads to launch TP workers
for sock in sockets:
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment