Skip to content

Commit e42f8de

Browse files
committed
update tp pp default
Signed-off-by: Maanu Grover <[email protected]>
1 parent 681ce76 commit e42f8de

File tree

1 file changed

+4
-4
lines changed

1 file changed

+4
-4
lines changed

megatron/training/checkpointing.py

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -1418,13 +1418,13 @@ def load_checkpoint(ddp_model, optimizer, opt_param_scheduler, load_arg='load',
14181418
ckpt_args = state_dict.get("args")
14191419

14201420
if not hasattr(ckpt_args, "tensor_model_parallel_size"):
1421-
print_rank_0("WARNING: TP size not found in checkpoint args, using 1 as default.")
1421+
print_rank_0("WARNING: TP size not found in checkpoint args, using 0 as default.")
14221422
if not hasattr(ckpt_args, "pipeline_model_parallel_size"):
1423-
print_rank_0("WARNING: PP size not found in checkpoint args, using 1 as default.")
1423+
print_rank_0("WARNING: PP size not found in checkpoint args, using 0 as default.")
14241424

14251425
ckpt_tp_pp = (
1426-
getattr(ckpt_args, "tensor_model_parallel_size", 1),
1427-
getattr(ckpt_args, "pipeline_model_parallel_size", 1),
1426+
getattr(ckpt_args, "tensor_model_parallel_size", 0),
1427+
getattr(ckpt_args, "pipeline_model_parallel_size", 0),
14281428
)
14291429
run_tp_pp = (
14301430
args.tensor_model_parallel_size,

0 commit comments

Comments
 (0)