Fixed bug in precision logic #232
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
My change here aligns val() with the impimentation in upstream. Without my change, the data samples will ALWAYS be cast to half when run on CUDA. However, the model may not be half precision in this case as per the logic above this section (The "if training: ... else: ...", specifically lines 145&149). This means the code ALWAYS crashes when run on CUDA when not passing --half as the model weights will be normal precision, but all input will be half precision. My change ensures that weights and input have same precision. Additionally, half precision is used if and only if --half is passed, the devices is not cpu, and the specific backend supports half precision
|
It may be worth mentioning that val.py will not crash if "--device cpu" is given, as that would make everything normal precision. This is the case for your test case of val, which is why your code has not been failing checks so far. |
This pull request was closed.
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Outdated suggestions cannot be applied.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.
My change here aligns val() with the impimentation in upstream. Without my change, the data samples will ALWAYS be cast to half when run on CUDA. However, the model may not be half precision in this case as per the logic above this section (The "if training: ... else: ...", specifically lines 145&149). This means the code ALWAYS crashes when run on CUDA when not passing --half as the model weights will be normal precision, but all input will be half precision.
My change ensures that weights and input have same precision. Additionally, half precision is used if and only if --half is passed, the devices is not cpu, and the specific backend supports half precision