Reproduce baseline with new Tracks dataset

feldberlin / wavenet

An unconditioned Wavenet implementation with fast generation.

3 stars 0 forks source link

Reproduce baseline with new Tracks dataset #32

Closed purzelrakete closed 2 years ago

purzelrakete commented 3 years ago

What

The tracks dataset was update to allow for overlapping of examples. Training should match the previous baseline.

Results

Write up the results of your experiment once it has completed and has been analysed. Include links to the treatment run, and also to the baseline if appropriate.

Acceptance Criteria

[x] Results match baseline

purzelrakete commented 3 years ago

Experiment

See the baseline in #18.

Treatment with overlapping examples maestro-1631824642:

 bin/train maestro -p batch_size 12 -p batch_norm True -p learning_rate 0.05 -p max_epochs 12

purzelrakete commented 3 years ago

Results

Reproduction experiment results:

	train nll	test nll
baseline	1.475	1.322
treatment	1.273	1.322

We can say that we could reproduce the baseline.