Daniel Ritchie
Results
2
comments of
Daniel Ritchie
With PyTorch trainer you can replace `trainer.train()` with `trainer.train(resume_from_checkpoint=)` _e.g. "checkpoint-1234"_ or `trainer.train(resume_from_checkpoint = True)` [More info from HF docs](https://huggingface.co/docs/transformers/main_classes/trainer#transformers.Trainer.train.resume_from_checkpoint)
The [starchat-alpha blog post](https://huggingface.co/blog/starchat-alpha) explains why you're not getting the results you are looking for (and also provides an example of their **StarChat** solution which I think would do what...