File 96beb97.diff of Package armnn.16481

Overview Repositories Revisions Requests Users Attributes Meta

File 96beb97.diff of Package armnn.16481

From 96beb97f0ce6650d64fb15d9c920d70a1572672c Mon Sep 17 00:00:00 2001
From: Keith Mok <ek9852@gmail.com>
Date: Thu, 07 Jan 2021 17:24:58 -0800
Subject: [PATCH] Adopt the change of tflite builtin code

This fixes tflite parser cannot parse model with new schema
Tensorflow 2.4 updated the schema to version 3.
And breaks armnn tensorflow lite parser.
Adopt changes to fix it.
But it will require using Tensorflow 2.4+ schema_generated.h

Signed-off-by: Keith Mok <ek9852@gmail.com>
Change-Id: I9eed8542e0882a39b7799a4e870ad09c69c7808f
---

diff --git a/src/armnnTfLiteParser/TfLiteParser.cpp b/src/armnnTfLiteParser/TfLiteParser.cpp
index 8e0fae6..183e337 100644
--- a/src/armnnTfLiteParser/TfLiteParser.cpp
+++ b/src/armnnTfLiteParser/TfLiteParser.cpp
@@ -645,7 +645,8 @@
             for (OperatorPtr const& op : subgraph->operators)
             {
                 auto const& opCodePtr = m_Model->operator_codes[op->opcode_index];
-                auto builtinCode = opCodePtr->builtin_code;
+                auto builtinCode = std::max(opCodePtr->builtin_code,
+                        static_cast<tflite::BuiltinOperator>(opCodePtr->deprecated_builtin_code));
 
                 if (builtinCode > tflite::BuiltinOperator_MAX)
                 {
@@ -765,7 +766,8 @@
     const auto & operatorPtr = m_Model->subgraphs[subgraphIndex]->operators[operatorIndex];
 
     auto opcodeIndex = operatorPtr->opcode_index;
-    auto opcode      = m_Model->operator_codes[opcodeIndex]->builtin_code;
+    auto opcode      = std::max(m_Model->operator_codes[opcodeIndex]->builtin_code,
+            static_cast<tflite::BuiltinOperator>(m_Model->operator_codes[opcodeIndex]->deprecated_builtin_code));
 
     if (!m_Options || !m_Options.value().m_StandInLayerForUnsupported)
     {

Places

File 96beb97.diff of Package armnn.16481

Places