Pp/cpc ctc stats #5
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…ignments) for specidied "capture DS"
janchorowski
approved these changes
Mar 6, 2021
cpc/criterion/soft_align.py
Outdated
| captureRes = {} | ||
| if 'pred' in captureOptions: | ||
| # need to remove 'predict self loop' stuff from captured data | ||
| captureRes['pred'] = predictions[:,:,:,1:] |
There was a problem hiding this comment.
I am unsure about this [1:] at the end - is it to skip the self-loop?
There was a problem hiding this comment.
Yeah, I was removing it as it's not really a prediction, but forgot it would spoil alignment indices - fixed, now self loop is also saved in predictions
|
Yeah, another function seems better - I just didn't want to have entirely separate run file with constructing all stuff from args and went a bit too far in "everything together" direction; updated |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Added option to capture data for future visualizations (save them across different epochs during training, or just see what is produced after the training) - representations, CPC predictions and CPC-CTC alignments
Example run that saves data each 2 epochs for 33% of provided 'capture dataset':
python train.py --pathDB /pio/scratch/1/i283340/MGR/zs/ds2
--pathTrain /pio/scratch/1/i283340/MGR/zs/sometries/ds2part.txt
--pathVal /pio/scratch/1/i283340/MGR/zs/sometries/ds2part.txt
--pathCaptureDS /pio/scratch/1/i283340/MGR/zs/sometries/ds2part.txt
--captureDSfreq 33 --captureEachEpochs 2
--pathCaptureSave /pio/scratch/1/i283340/MGR/zs/capture/try2
--saveRepr --savePred --saveAlign
--pathCheckpoint /pio/scratch/1/i283340/MGR/zs/checkpoints/cpcctc_tests2
--file_extension .flac --n_process_loader 2 --max_size_loaded 40000000
--batchSizeGPU 16 --nPredicts 8 --CPCCTC --CPCCTCNumMatched 12
--CPCCTCSelfLoop --CPCCTCSkipBeg 1 --CPCCTCSkipEnd 2