We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
1 parent eb73662 commit c0312efCopy full SHA for c0312ef
1 file changed
modelopt/torch/distill/plugins/megatron.py
@@ -172,7 +172,7 @@ def _adjust_layer_index_for_pp(submodule_name, model_cfg):
172
if new_layer_idx < 0:
173
raise ValueError(f"Layer {submodule_name} does not fall on final PP rank.")
174
175
- new_submodule_name = submodule_name.replace(match.group(0), str(new_layer_idx))
+ new_submodule_name = submodule_name.replace(f".{match.group(0)}", f".{new_layer_idx}")
176
if parallel_state.get_tensor_and_context_parallel_rank() == 0:
177
logger.info(
178
f'Distillation: Renamed layer "{submodule_name}" on final PP rank to "{new_submodule_name}"'
0 commit comments