From b61cdc3b5174bbf40885f4acd1bde14981d11544 Mon Sep 17 00:00:00 2001
From: Scott Wolchok <swolchok@meta.com>
Date: Mon, 4 Mar 2024 16:49:12 -0800
Subject: [PATCH] Fix static runtime sigrid_hash precomputed multiplier pass

Reviewed By: pls331, houseroad

Differential Revision: D54336561
---
 torch_glow/src/ShapeInferenceEngine.cpp | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/torch_glow/src/ShapeInferenceEngine.cpp b/torch_glow/src/ShapeInferenceEngine.cpp
index 24dd73f0aa..81b13e78d5 100644
--- a/torch_glow/src/ShapeInferenceEngine.cpp
+++ b/torch_glow/src/ShapeInferenceEngine.cpp
@@ -3437,7 +3437,8 @@ ShapeInferenceEngine::argmin(const MetaStack &variableMetas) {
  *    int salt,
  *    int maxValue,
  *    Tensor multiplier_shift,
- *    bool hashIntoInt32
+ *    bool hashIntoInt32,
+ *    bool? noHashNegSalt
  * ) -> Tensor
  *
  *
@@ -3445,8 +3446,8 @@ ShapeInferenceEngine::argmin(const MetaStack &variableMetas) {
 Expected<TensorOutput>
 ShapeInferenceEngine::sigridHashPrecompute(const MetaStack &variableMetas) {
   RETURN_ERR_IF_NOT(
-      variableMetas.size() == 5,
-      strFormat("Expected 5 inputs, got %zu", variableMetas.size()));
+      variableMetas.size() == 6,
+      strFormat("Expected 6 inputs, got %zu", variableMetas.size()));
 
   TensorShape shape = variableMetas[0].shape<TensorShape>();