Safetensors, really?
#2
by
Bitsy
- opened
I get that you want to be secure but why not just a pytorch_model.bin files(s) ? including those would be really helpful. Thanks.
because the binaries eat up a lot of storage and i hate dealing with them.
hakurei
changed discussion status to
closed
Just get good foo
You can't just win an argument by closing the discussion! I want pickles! :cucumber:
turn it into a pickle yourself, i'm not throwing multiple several gb binaries around.
from transformers import AutoModelForCausalLM
AutoModelForCausalLM.from_pretrained('model').save_pretrained('model')
from transformers import AutoModelForCausalLM AutoModelForCausalLM.from_pretrained('model').save_pretrained('model')
Wut? doesn't work :/ please share your pickles they are tastey and actually usable.
safetensors are useable yes