I use this set --gradient_checkpointing \

[LeoLopezt]

But it returns the error
RuntimeError: The size of tensor a (1024) must match the size of tensor b (256) at non-singleton dimension 1.
Can you help me deal with this problem? Thanks

[xdobetter]

refer to #10 (comment)

[LeoLopezt]

If I want this set "--gradient_checkpointing " to be used, is there any other ways to deal with

[LeoLopezt]

cause a 24GB GPU isn't enough to train

[xdobetter]

By employing the methodologies outlined therein #10 (comment) , it should be possible to conduct training with a size below 24GB. Perhaps you might consider giving it a try.