Significant memory reduction through TCNN's new memory arena #128
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
Tom94
changed the title
fnysalehi
pushed a commit
to fnysalehi/instant-ngp-rendering
that referenced
this pull request
May 14, 2024
Significant memory reduction through TCNN's new memory arena
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Outdated suggestions cannot be applied.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.
The
foxscene now requires only 2.3 GB on an RTX 3090 w/FullyFusedMLP.Even when using Fp32 +
CutlassMLP(the worst case for memory usage),foxandlego/transforms_train.jsonfit into 8gb.Also fixes temporarily broken input gradients.
GPU Memory Arena
tiny-cuda-nn now has a memory arena that can be used for efficient per-stream allocation of temporary memory. As a consequence, much of the memory that previously needed to be pre-allocated can now be allocated and released on the fly (effectively re-using it across sequential computations). This ends up reducing memory usage by over 50% in many cases. See also #36