Comments (7)
I'm thinking that we can change the assert here to a warning based on a floating point close-equals that prints out a message to the user.
from persephone.
Thanks for the report, I'll get this fixed soon
from persephone.
assert test_per == test_ler
the assertion being failed on line 293 of model.py
from persephone.
Thanks for opening this issue.
@oadams I'd appreciate a comment about the intent behind the code that contains the assertion here:
persephone/persephone/model.py
Lines 293 to 296 in b4ce92f
from persephone.
The assertion would have been to ensure that the tensorflow label error rate was consistent with what was measured by my implementation of minimum edit distance at some point. But it looks like utils.batch_per() is using the NLTK implementation of edit distance anyway. It could be a floating point precision thing, but I'd like to see what the actual values are since if that's the case it's a bit surprising this hasn't popped up so far.
Either way, the assert should probably go since it doesn't help anyone.
from persephone.
It could be a floating point precision thing, but I'd like to see what the actual values are since if that's the case it's a bit surprising this hasn't popped up so far.
Yup! Good intuition. I ran the training again and had it print the two variables just before the assertion:
test_per= 0.2878205128205128 test_ler= 0.28782052
I've also got it paused on the debugger on the assertion line if you guys want me to explore further, please let me know soon. For the record i've hit this bug 3/3 times i've run the training, so it seems pretty reproducible.
from persephone.
@tensorfoo Hopefully the changes I made in #224 resolve your issue, if not please let us know
from persephone.
Related Issues (20)
- OpenFST dependency HOT 2
- adding word boundaries to the acoustic model for Na HOT 1
- Supply information about supported label types and feature types
- Segmentation of characters with unicode space characters HOT 2
- Python 3.7 support HOT 1
- Create a new release HOT 3
- Link to papers in the docs
- CorpusReader expects number of training examples to be divisible by batch size
- Link for Na example data is broken HOT 2
- Stable hosting (+long-term archiving) of preprocessed data sets HOT 1
- Stable hosting (+long-term archiving) of generated models (acoustic models etc) HOT 4
- Making use of audio chunks of more than 10 seconds HOT 2
- harmonize license between Persephone and Elpis? HOT 11
- Suggestions wanted with training parameters HOT 13
- Missing Artwork HOT 2
- Integrating Persephone into a production ML pipeline HOT 3
- GPU support? HOT 7
- Feature type fbank_and_pitch throws error HOT 10
- Update documentation for Kaldi path configuration
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from persephone.