Compute capability < 7.5 detected!

[shon-otmazgin]

I trying to run:

from transformers import AutoModelForSeq2SeqLM
from peft import get_peft_model, LoraConfig
model_name_or_path = "facebook/opt-13b"

peft_config = LoraConfig(
    task_type="CAUSAL_LM", inference_mode=False, r=64, lora_alpha=32, lora_dropout=0.1
)

model = AutoModelForSeq2SeqLM.from_pretrained(model_name_or_path)
print(model.print_trainable_parameters())
model = get_peft_model(model, peft_config)
print(model.print_trainable_parameters())

but then I getting:
Compute capability < 7.5 detected! Only slow 8-bit matmul is supported for your GPU!"
NameError: name 'cuda_setup' is not defined

I understood that this error comes from bitsandbytes so I found this issue:
bitsandbytes-foundation/bitsandbytes#124

My hardware is tesla v100-sxm2-32gb (Volta)

It is possible to run PEFT on my hardware? I don't really need int8, fp16 also should be good.

this is related also to my attempts to FT a large LM on my hardware:
https://discuss.huggingface.co/t/finetune-llm-with-deepspeed/31589

Thanks,
Shon

[younesbelkada]

Hi @shon-otmazgin
I think if you don't need bnb you can just uninstall bitsandbytes and the error should disappear

[djaym7]

bitsandbytes 0.35.0 solves this but start another issue :
Traceback (most recent call last):
File "train_full_csv_int8Training.py", line 463, in
train(locals(),train_inputs,subtopics_train_targets,val_inputs,subtopics_val_targets)
File "train_full_csv_int8Training.py", line 400, in train
trainer = Trainer(
File "/usr/local/lib/python3.8/dist-packages/transformers/trainer.py", line 382, in init
raise ValueError(
ValueError: The model you want to train is loaded in 8-bit precision. if you want to fine-tune an 8-bit model, please make sure that you have installed bitsandbytes>=0.37.0

[djaym7]

bitsandbytes-foundation/bitsandbytes#179

[younesbelkada]

@djaym7 can you try with bitsandbytes==0.37.0 ?

[djaym7]

No can't bitsandbytes-foundation/bitsandbytes#179

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.