Mike Izbicki

GPU work is something that I've been very interested in, but haven't really looked at yet at all. I _think_ the way the linear algebra stuff in subhask is structured...

I've merged #71. Does that fix the problem?

> Well, yes, but that set subhask to an old commit well behind master. Was that your intention? Yes. I often make updates to subhask without testing them on HLearn....

I have a short and long answer for what I think is the problem for you. **The short answer:** if you add all the parameters into the command line that...

65% accuracy is pretty low for this problem. I get 78% with the command ``` $ python3 names.py --train --gradient_clipping --model=gru --learning_rate=1e-1 --batch_size=10 --hidden_layer_size=128 --num_layers=1 --samples=100000 ``` And then by...

My wife has a doctor's appointment during our office hours tomorrow morning (Monday, 6 April), and so I need to reschedule the office hours to 2-3pm. If you can't make...

My initial training is done with ``` $ python3 names.py --train --model=gru --learning_rate=1e-1 --batch_size=10 --gradient_clipping --hidden_layer_size=128 --num_layers=1 --num_samples=100000 ``` and then I decay the learning rate by 10 twice.

Hmm... I may have to look at this in person with you. At a high level, there's two possibilities: 1. You're not setting the `scores` vector correctly. To verify this,...

It looks to me like your blue model would achieve >= 0.8 accuracy if you let it run for the full 100k samples instead of stopping it after 30k. It's...

@zhh1997zhh Ahh... I think I see the problem now. **short answer:** Your batch size of 1 is very small. I think if you change the batch size to 10 than...