-
Notifications
You must be signed in to change notification settings - Fork 31k
Closed
Description
System Info
Traceback (most recent call last):
File "/workspace/run/run_llm.py", line 717, in <module>
main()
File "/workspace/run/run_llm.py", line 644, in main
trainer = Trainer(
File "/opt/conda/lib/python3.10/site-packages/transformers/trainer.py", line 342, in __init__
self.create_accelerator_and_postprocess()
File "/opt/conda/lib/python3.10/site-packages/transformers/trainer.py", line 3900, in create_accelerator_and_postprocess
"activation_checkpointing", fsdp_plugin.activation_checkpointing
AttributeError: 'FullyShardedDataParallelPlugin' object has no attribute 'activation_checkpointing'
transformers/src/transformers/trainer.py
Lines 3893 to 3907 in aea7614
| # post accelerator creation setup | |
| if self.is_fsdp_enabled: | |
| fsdp_plugin = self.accelerator.state.fsdp_plugin | |
| fsdp_plugin.limit_all_gathers = self.args.fsdp_config.get( | |
| "limit_all_gathers", fsdp_plugin.limit_all_gathers | |
| ) | |
| fsdp_plugin.activation_checkpointing = self.args.fsdp_config.get( | |
| "activation_checkpointing", fsdp_plugin.activation_checkpointing | |
| ) | |
| if fsdp_plugin.activation_checkpointing and self.args.gradient_checkpointing: | |
| raise ValueError( | |
| "The activation_checkpointing in FSDP config and the gradient_checkpointing in training arg " | |
| "can't be set to True simultaneously. Please use FSDP's activation_checkpointing logic " | |
| "when using FSDP." | |
| ) |
The 'FullyShardedDataParallelPlugin' class in accelerate version v0.22.0 does not have 'activation_checkpointing'. but the main branch does.
Who can help?
No response
Information
- The official example scripts
- My own modified scripts
Tasks
- An officially supported task in the
examplesfolder (such as GLUE/SQuAD, ...) - My own task or dataset (give details below)
Reproduction
Expected behavior
s1ghhh and xhluca
Metadata
Metadata
Assignees
Labels
No labels