add e-branchformer result for tedlium3 and add checker for text output length #5130
Conversation
Codecov Report
@@ Coverage Diff @@
## master #5130 +/- ##
==========================================
- Coverage 74.99% 74.99% -0.01%
==========================================
Files 618 618
Lines 55586 55588 +2
==========================================
+ Hits 41688 41689 +1
- Misses 13898 13899 +1
Flags with carried forward coverage won't be shown. Click here to find out more.
📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more |
|
Thanks!
|
|
@Some-random, any update? |
for more information, see https://pre-commit.ci
So sorry for the delay, so how I missed the feedback on this PR. I've put the results from E-Branchformer in the front and added links for model on HuggingFace. |
| if len(text_ints) > 100: | ||
| logging.warning( | ||
| "The length of the text output exceeds 100, " | ||
| "which may cause OOM on the GPU." | ||
| "Please ensure that the data processing is correct and verify it." | ||
| ) |
There was a problem hiding this comment.
@sw005320 @Some-random Sorry, really late on that (I didn't sync w/ master in a while) but why was it added and how was the limit defined?
I was a bit confused when I saw that thousand of warnings were displayed in my training logs!
There was a problem hiding this comment.
Yes, it was added to a warning of the large memory consumption, but 100 is probably too small.
So, we can change it to 500 or more.
Sorry about it.
There was a problem hiding this comment.
Thanks and no worries, I'll open a PR!
Edit: You're too fast for me Professor! 😄
No description provided.