update LlamaPreTrainedModel for gradient checkpointing refactor
Browse files- modeling_llama_yarn.py +0 -4
modeling_llama_yarn.py
CHANGED
@@ -860,10 +860,6 @@ class LlamaPreTrainedModel(PreTrainedModel):
|
|
860 |
if module.padding_idx is not None:
|
861 |
module.weight.data[module.padding_idx].zero_()
|
862 |
|
863 |
-
def _set_gradient_checkpointing(self, module, value=False):
|
864 |
-
if isinstance(module, LlamaModel):
|
865 |
-
module.gradient_checkpointing = value
|
866 |
-
|
867 |
|
868 |
LLAMA_INPUTS_DOCSTRING = r"""
|
869 |
Args:
|
|
|
860 |
if module.padding_idx is not None:
|
861 |
module.weight.data[module.padding_idx].zero_()
|
862 |
|
|
|
|
|
|
|
|
|
863 |
|
864 |
LLAMA_INPUTS_DOCSTRING = r"""
|
865 |
Args:
|