Hi
I have trained a model with Pytorch lightning and have a ckpt.pth checkpoint file.
I would now like to make this available as a model on huggingface to run inference tests.
I have a script on my machine and cloud GPUs that allows me to run inference with the checkpoint file, but I am really struggling with the next steps as to how to prepare it for huggingface.
I did manage to make a .bin version, upload to HF and do some tests - but the outputs were garbled compared to what I get when run inference on my cloud gpus.
This is a pytorch model, so not trained via the HuggingFace libraries?
I have searched low and high - but really don’t know what to do next…
Sincerest thanks