correct the tp_plan logic (#8850)
This commit is contained in:
@@ -211,16 +211,13 @@ class TransformersForCausalLM(nn.Module):
|
||||
Apply the model's tensor parallelization plan.
|
||||
Currently only supports linear layers.
|
||||
"""
|
||||
if not self.model.supports_tp_plan:
|
||||
if tp_size <= 1:
|
||||
return
|
||||
tp_plan = getattr(self.model.config, "base_model_tp_plan", None) or {}
|
||||
|
||||
if not tp_plan and self.tp_size > 1:
|
||||
raise ValueError(
|
||||
f"{type(self.model)} does not support tensor parallel yet!"
|
||||
)
|
||||
|
||||
tp_plan = getattr(self.model.config, "base_model_tp_plan", None) or {}
|
||||
|
||||
def _tensor_parallel(module: nn.Module, prefix: str = ""):
|
||||
for child_name, child_module in module.named_children():
|
||||
qual_name = maybe_prefix(prefix, child_name)
|
||||
|
||||
Reference in New Issue
Block a user