You signed in with another tab or window.Reload to refresh your session.You signed out in another tab or window.Reload to refresh your session.You switched accounts on another tab or window.Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: py/torch_tensorrt/dynamo/conversion/impl/normalization/ops.py
+3-2Lines changed: 3 additions & 2 deletions
Original file line number
Diff line number
Diff line change
@@ -551,8 +551,9 @@ def cdist_forward(
551
551
- For x1.shape = [10, 5] (10 vectors, each of 5 features) and x2.shape = [20, 5] (20 vectors, each of 5 features),
552
552
since there are no batch dimensions to match, the output shape is simply [10, 20], comparing all vectors from x1 against all vectors from x2.
553
553
554
-
Note: The `compute_mode` parameter is designed to optimize the performance of the Euclidean distance calculation, especially useful when working with large datasets.
555
-
This parameter allows you to control how the distances are computed, with different modes available to leverage matrix multiplication for speed improvements.
554
+
Note: The `compute_mode` parameter is designed to optimize the performance of the Euclidean distance calculation,
555
+
especially useful when working with large datasets. This parameter allows you to control how the distances are computed,
556
+
with different modes available to leverage matrix multiplication for speed improvements.