Commit 95a72249 authored by Doe Hyun Yoon's avatar Doe Hyun Yoon Committed by TensorFlower Gardener
Browse files

Fix the condition in ShouldUpdateOutputValue() in static shape inference. 

When using aggressive_shape_inference, running EvaluateNode() takes too much resources
(compute and memory) if input and output tensors are large. Previously, set 17 as
threshold for small / large tensors, but the condition check logic was incorrect.

This CL fixes that and avoid costly EvaluateNode() in case output tensor are
huge (e.g., Fill op that takes a shape describing large tensors).

PiperOrigin-RevId: 232493327
parent de9f92ce
Show whitespace changes
Inline Side-by-side
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Please to comment