feat: added slice onnx import #1856
Conversation
JachymPutta
force-pushed
the
jp/slice_onnx_import
branch
from
959c74a to
41c821d
Compare
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #1856 +/- ##
==========================================
- Coverage 86.30% 86.10% -0.20%
==========================================
Files 774 777 +3
Lines 90072 90649 +577
==========================================
+ Hits 77733 78056 +323
- Misses 12339 12593 +254 ☔ View full report in Codecov by Sentry. |
| let input = Tensor::<Backend, 2>::from_floats( | ||
| [ | ||
| [1., 2., 3., 4., 5., 6., 7., 8., 9., 10.], | ||
| [1., 2., 3., 4., 5., 6., 7., 8., 9., 10.], |
There was a problem hiding this comment.
Maybe change the values of the second dim just to make it clear what data is sliced with this test 😅
| // compine starts and ends into ranges | ||
| // example: starts = [0, 0, 0, 0], ends = [1, 1, 1, 1] | ||
| // ranges = [0..1, 0..1, 0..1, 0..1] | ||
| let starts = &self.starts; | ||
| let ends = &self.ends; |
There was a problem hiding this comment.
I think we still need to check the axes since there is no guarantee that they always including the leading axes (e.g., if axes = [2, 3] for values of starts and ends), but it's not entirely clear even as described in the spec.
If we keep this solution without checking for non leading axes, I would at least parse the values of axes and error on invalid/unsupported values (i.e., skipping leading axes).
Checklist
run-checks allscript has been executed.Related Issues/PRs
Helping with #1714
Changes
Added the slice import to onnx
Testing
cargo test -p burn-import --color=always -- --color=alwayscargo test -p onnx-tests --color=always