Added Temporal Score functions #72
Conversation
…ertSpaceCharacterPerturbations and added test cases
Codecov Report
@@ Coverage Diff @@
## master #72 +/- ##
==========================================
+ Coverage 94.91% 95.27% +0.36%
==========================================
Files 8 9 +1
Lines 334 360 +26
==========================================
+ Hits 317 343 +26
Misses 17 17
Continue to review full report at Codecov.
|
tests/test_temporal_score.py
Outdated
| import json | ||
| from pathlib import Path | ||
|
|
||
| json_path = Path("decepticonlp/metrics/glove_sample.json") |
There was a problem hiding this comment.
Shouldn't the glove file be in tests directory?
|
|
||
| class RankCharacters: | ||
| """ | ||
| Accepts a feature vector torch tensor and outputs a temporal ranking of characters |
There was a problem hiding this comment.
Can we have proper docstrings here?
| def temporal_score(self, sentence): | ||
| """ | ||
| Considering a input sequence x1,x2,...,xn | ||
| we calculate T(xi) = F(x1,x2,...,xi) - F(x1,x2,...,xi-1) |
There was a problem hiding this comment.
Docstrings in our preferred format needed here too? Including information about params and return values?
|
|
||
| def temportal_tail_score(self, sentence): | ||
| """ | ||
| Considering a input sequence x1,x2,...,xn |
|
|
||
| def combined_score(self, sentence, lambda_): | ||
| """ | ||
| Computes combined temporal_score, temportal_tail_score |
There was a problem hiding this comment.
I'll add all of the docstrings
| import torch.nn as nn | ||
|
|
||
|
|
||
| class RankCharacters: |
There was a problem hiding this comment.
Is it ranking characters or words?
There was a problem hiding this comment.
My bad, this is rank characters. Before this I did it on a character level but later changed it to a word level, forgot to change the class name :P. I'll change this rn.
| inputs = sentence.reshape(len(sentence), 1, -1) | ||
| model = nn.RNN(inputs.shape[2], 1) | ||
| with torch.no_grad(): | ||
| _, pred = model(inputs) |
There was a problem hiding this comment.
Shouldn't the model be pre-trained on the task? I mean user should have an option to use his own models to test this
There was a problem hiding this comment.
@rajaswa , this is the temporal scorer, it selects words based on highest decrease in rnn outputs.
There was a problem hiding this comment.
- I am aware of that. But monitoring output from an untrained RNN doesn't make sense at all
- This should be usable by all models working with text sequences
There was a problem hiding this comment.
Cool I'll make the changes in that way then !
|
Kudos, SonarCloud Quality Gate passed!
|
No description provided.