Alex Dmitrewski Manager   •   about 3 years ago

Computer vision questions so far

Hi everyone, I'll just reply here to a few questions I've had so far.

1) While we care about the final image and the quality, please invest in the whole process. For example, show us that your model fine-tuning worked (e.g., 4 before images vs 4 post-training images), and how you chose your parameters.

2) Don't be afraid of including aspects of the sample notebooks in your solutions - we expect you to expand on them! So you could use similar image ideas, but maybe you could then fine-tune your stage 2 model.

3) You can save fine-tuned model weights locally, but also to HuggingFace for later loading.

Also I I think it is mentioned in the notebook but you do need to create a HuggingFace account!

Alex