Since some large models could take hours to train, do we have to give an option to train the model in run.py, or is loading a trained model sufficient ?
Also, even if we want to "simply" import the trained model in run.py, we have a problem because the model doesn't fit on git (600Mo - 1Go). How can we do ? Also, if we try a lot of different embeddings, the total size of our repo becomes very big.
run.py train or load trained model
Hello,
Regarding the text-classification project :
Since some large models could take hours to train, do we have to give an option to train the model in run.py, or is loading a trained model sufficient ?
Best regards
2
I would as well be interested in this answer for the road-segmentation project.
Also, even if we want to "simply" import the trained model in run.py, we have a problem because the model doesn't fit on git (600Mo - 1Go). How can we do ? Also, if we try a lot of different embeddings, the total size of our repo becomes very big.
Add comment