Merge test branch into main #26
Merged
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
- Minor Change is made test_pytorch_ref.c file so that it works with current ci, as string matching from the artificat take place and hence it should be exact same. - check_all_results.py is changed to sorted output for consistent ordering -run-tests.yml is also slightly changed to as in the Windows job, the path to the executable is incorrect, which is rectified - LOCALLY TESTED
- Add, matmul, mean, mul, mulf, sub, sum
…r partipular operator under particular subtest category
Add Test Suite for cTensor
- Once the reduced dimension is found, we "unsqueeze" the upstream gradient (e.g., from {2} to {2, 1}), making it ready for broadcasting. This is a clean, zero-copy operation that solves the shape mismatch elegantly.
- fixed a bug in GradFn_mean where it was incorrectly calculating the gradient value based on the output tensor's size instead of the input's - It now correctly computes the 1/N gradient.
Fix: Stabilize Backpropagation for Sum and Mean
This reverts commit 6332e53.
Add: autograd support for min and max tensor operations
- Refactored GradFn_div and GradFn_pow to handle broadcasting correctly with modulo indexing - Preserved original tensors in Tensor_div and Tensor_pow before broadcasting - Fixed gradient computation in both operations to work with tensors of different shapes - Simplified gradient function implementation by reusing result tensor - Fixed power operation's gradient calculation to use cached result instead of recalculating
[FIX] Broadcasting in division and power operations with tests
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Outdated suggestions cannot be applied.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 current version supports:
add, div, matmul, max, min, mean, mul, mulf, pow, reciprocal, square, sub, sum. Both forward and backward operations are supported.sum, mean, min, maxcan take an extra argument to reduce the specific dimension like pytorch.linear, relu./src2/main.c