Positional Encoding Shape Mismatch in Clay Encoder #377
Conversation
This file contains hidden or 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
neerajk
force-pushed
the
main
branch
2 times, most recently
from
6cd081d to
201dde7
Compare
for more information, see https://pre-commit.ci
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.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.
RuntimeError: The size of tensor a (1024) must match the size of tensor b (1020) at non-singleton dimension 2
Root Cause
The error was caused by incorrect dimensionality in the positional metadata encoding. Specifically:
The time and latlon tensors were incorrectly shaped as [B, 2] instead of [B, 4], resulting in a combined metadata tensor of shape [B, L, 8] instead of the expected [B, L, 12].
This led to a mismatch when adding positional encodings to the patch embeddings, which are shaped [B, L, 1024].
Fixes Applied
Corrected the normalize_timestamp() and normalize_latlon() functions to return 4 components each: sin/cos(week/hour) and sin/cos(lat/lon) respectively.
Ensured the metadata tensors are properly stacked and repeated to match the patch sequence length (L).
Verified that the final positional encoding tensor matches the expected shape [B, L, 1024] before addition.
Validation
Encoder now runs successfully with wall-to-wall Sentinel-2 inputs.