Supporting gradient checkpointing for QLORA
#16
by
ospanbatyr
- opened
Hi everyone,
While trying to finetune OLMo-7B with QLORA, OLMoForCausalLM
does not support gradient checkpointing error is thrown in the prepare_model_for_kbit_training(model)
line. Traceback:
Traceback (most recent call last):
File "/scratch/users/oince22/hpc_run/CartographyFT/src/driver.py", line 39, in
main
run_main(P, logger)
File "/scratch/users/oince22/hpc_run/CartographyFT/src/driver.py", line 68, in
run_main
llm, tokenizer = P.get_lm()
^^^^^^^^^^
File "/scratch/users/oince22/hpc_run/CartographyFT/src/params.py", line 311,
in get_lm
model = prepare_model_for_kbit_training(model)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File
"/kuacc/users/oince22/.conda/envs/icl/lib/python3.11/site-packages/peft/utils/ot
her.py", line 139, in prepare_model_for_kbit_training
model.gradient_checkpointing_enable(**gc_enable_kwargs)
File
"/kuacc/users/oince22/.conda/envs/icl/lib/python3.11/site-packages/transformers/
modeling_utils.py", line 2092, in gradient_checkpointing_enable
raise ValueError(f"{self.__class__.__name__} does not support gradient
checkpointing.")
ValueError: OLMoForCausalLM does not support gradient checkpointing.
Hello @ospanbatyr , did you resolve this issue? If yes, I will close this.
amanrangapur
changed discussion status to
closed