upgrade nerfacc to 0.3.x with CUB accel #1187
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
|
The quality is numerically the same as before. |
tancik
approved these changes
Dec 30, 2022
This was
linked to
issues
Dec 30, 2022
Closed
tancik
pushed a commit
to dozeri83/nerfstudio
that referenced
this pull request
Jan 20, 2023
lucasthahn
pushed a commit
to tne-ai/nerfstudio
that referenced
this pull request
Jan 26, 2023
chris838
pushed a commit
to chris838/nerfstudio
that referenced
this pull request
Apr 22, 2023
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Outdated suggestions cannot be applied.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.
White -- nerfacc 0.3.x.
Blue -- nerfacc 0.2.x.
nerfacc 0.3.x uses CUB to accelerate the accumulation (both fwd and bwd) when available (cuda toolkit >= 11.5). Test rays / secs got ~10% speed up. For some reason the training speed is slightly slower with nerfacc 0.3.x, which needs to be further investigated.
Note the CUB acceleration is more noticeable when the #samples per ray is less, so the gap for test rays / secs got larger when converge. But it's unclear what's the reason of the slow down during training.