Size mismatch error when loading Mistral-Small

#1
by Srirakshith - opened

I'm trying to load the 4 bit quantised model with an L4 GPU, and it keeps throwing up the following error.

ValueError: Trying to set a tensor of shape torch.Size([83886080, 1]) in "weight" (which has shape torch.Size([5120, 32768])), this looks incorrect.

Is there a setting that I need to change to overcome this?

Sign up or log in to comment