MycroftAI
/
mycroft-precise
mirror of https://github.com/MycroftAI/mycroft-precise.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

Don't train when activation subsides in incremental training 

This allows batching up larger amounts of data before retraining
pull/10/head
Matthew D. Scholefield 2018-04-16 17:31:53 -05:00
parent 607db0437f
commit 9418990ff1
1 changed files with 1 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
precise/scripts/train_incremental.py
View File

@ -37,7 +37,7 @@ usage = '''
Number of epochs to train before continuing evaluation
:-ds --delay-samples int 10
Number of timesteps of false activations to save before re-training
Number of false activations to save before re-training
:-c --chunk-size int 2048
Number of samples between testing the neural network
@ -137,8 +137,6 @@ class IncrementalTrainer:
save_audio(name, audio_buffer)
print()
print('Saved to:', name)
elif samples_since_train > 0:
samples_since_train = self.args.delay_samples
if not save_test and samples_since_train >= self.args.delay_samples and self.args.epochs > 0:
samples_since_train = 0