Closed griff4692 closed 1 year ago
It looks like you can run
for inference. I can't find example call args in the docs, however
Hey @griff4692 sorry for the late response, this is definitely a bug with the predict API. We'll be getting rid of the CLI soon, so will also need to update the docs!
@SeanNaren, I believe that we have finished the API cleaning, so now, just to polish the docs? cc: @rohitgr7
🐛 Bug
Running this, you get a device mismatch since it puts the model inputs on CPU. However, I looked at the pipeline docs on HF and tried passing
device='cuda:0
to model.hf_predict yet I get the following error:Any advice? This seems like a pretty standard use case so I think there should be an easy fix.
Thanks