Mike Izbicki

The key idea is that the `input_ids` variable is what needs to get passed into to the bert model. So base on the code you have, that's what should be...

You need to fix ensure that your versions of pytorch and tensorboard are compatible. See #40 for the versions I used.

@nikpap123 My guess is that you didn't actually fix the versions, and have some old versions still around. Lot's of people have run into this issue on the assignment, and...

Yes, BERT is about 100x slower than the other models we have used. Training the model to convergence will take several days on a normal laptop. Fortunately, you are not...

I'm on office hours right now.

@raequan Sorry, I had to leave for another meeting before you showed up. @raequan @benfig1127 I'll have office hours tomorrow morning at 9am.

1. Any smoothing value is fine. 0.99 would be a good choice. 1. You only need a single run. There is no need for warm starting. 1. Your loss value...

You'll want to use tsne, and you'll need to adjust the hyperparameters until you get some clusters forming.

I think you could probably already have clusters if you use the tsne algorithm with the right paramters.

The filename should be `corona.multilang100.jsonl.gz`, but your filename is `corona.multilang100.jsonl`. So it looks like you're right that you've unziped the file. Running ``` $ gzip corona.multilang100.jsonl ``` will rezip the...