[FIX] ADR model size about yoruba-adr HOT 3 CLOSED

Learning more about the Pytorch model structure, we see that the optimizer’s state_dict is also saved as it contains state and parameters that are updated as the model trains. So this is important for checkpoint models, to resume training, but it is NOT important for the final model, so we can delete this part of the model.

If you trained your model using Adam, you need to save the optimizer state 
dict as well and reload that. Also, if you used any learning rate decay, you need 
to reload the state of the scheduler because it gets reset if you don’t, and you 
may end up with a higher learning rate that will make the solution state oscillate. 
Finally, if you have any dropout or batch norm in your model architecture, and 
you saved your model after a test loop (in which case model.eval() was called), 
make sure to call model.train() before the training loop.

from yoruba-adr.

Fixed in 5781370, new model is 67MB which is still bigger than the recommended 50MB, but at least it still pushes. I should keep track of the history and other sources of bloat in this repo.

from yoruba-adr.

An alternative if you want to use big files on the web, but host elsewhere than somewhere easily curl-able or within the github repo itself. Appreciate the fiddly-ness:

https://medium.freecodecamp.org/how-to-transfer-large-files-to-google-colab-and-remote-jupyter-notebooks-26ca252892fa

from yoruba-adr.