Is my training running?

[Quambus]

Hi everyone, I'm a complete noob with Python/Torch, but I thought this whole thing was really cool and decided to give it a try. I've managed to process my data and install Torch with all the other plugins. However, when I try train.lua and then sample.lua, its giving me nothing. My file name was a.h5 and a.json. Could anyone please help? Thanks in advance.

[ChrisCummins]

Two things I'd check:

1. Is your system doing something? If you are using CUDA then you can check nvidia-smi, else use top.
2. Does it work if you pre-process and train on the included data/tiny-shakespeare.txt dataset?

[Quambus]

I'm not using CUDA, and I processed the Shakespeare data set, but when I try to sample it, its still not giving me anything. Thanks again

[ChrisCummins]

Did you train on Shakespeare before sampling?

[Quambus]

I tried, and this is what I'm getting

[antihutka]

You're typing the commands into Torch's shell. You need to type them into your OS's shell.

[Quambus]

Ah, I'm doing some janky setup where I'm running Bash on Ubuntu since I have a PC. Could that be the problem?

[HaraldKorneliussen]

No, bash is the standard on ubuntu and I don't know what you would be running it on if you didn't have a PC. Just instead of writing "th" to start torch, you write "th train.lua -input_h5 " etc.

[Quambus]

Oh okay, thanks! I tried that and got this error:

I think I'm in the wrong directory? How do I get into the right directory? Also, is there a specific place where I need to put my training files? Thanks

[one-new-message]

Do ls and see which directory you have torch-rnn in (Probably "torch-rnn"). Then, just cd torch-rnn or whatever directory it was. Now try training. train.lua is in the directory you just changed to.

Alternatively, you can append torch-rnn/(Or whichever directory it was) in front of train.lua, like this: th torch-rnn/train.lua -input_h5 tiny-shakespeare.h5 -input_json tiny-shakespeare.json

Whichever method works best for you.