ml5js / training-charRNN

Training charRNN model for ml5js
Other
96 stars 46 forks source link

Change preprocess files naming to avoid cache #19

Closed peilingjiang closed 4 years ago

peilingjiang commented 4 years ago

Avoid models trained all become the same as the first one when running a continues training script for multiple models. Also adjust hyperparameter names in README to make it easier to understand.

peilingjiang commented 4 years ago

18

shiffman commented 4 years ago

Fantastic, thank you for this @peilingjiang! Do you think it would make more sense to write these files out to a different directory and not the input one?

peilingjiang commented 4 years ago

@shiffman Thank you for pointing that out! I've changed the directory storing the processed files and now they'll be stored in a new folder in the root directory.

shiffman commented 4 years ago

Wonderful, thank you!