Different model size

#3
by seigishi - opened

In the files and versions tab, the model size is showing 545 MB but when I download using pipeline it is around 1GB. What is the difference ???

It depends on whether you download it in FP32, FP16 etc. See the discussion here: https://discuss.huggingface.co/t/why-is-uploaded-model-twice-the-size-of-actual-model/18782

The solution from the discussion is: "You can pass torch_dtype=torch.float16 to your call to from_pretrained to load the pretrained model in half precision, or convert your model to half precision before saving it."
(Note that last time I tried, mDeberta did not support FP16 though)

MoritzLaurer changed discussion status to closed

Sign up or log in to comment