Remove unnecessary option from convert_onnx_models_to_ort.py, fix old instructions. (#11088)

Remove unnecessary --nnapi_partitioning_stop_ops option from convert_onnx_models_to_ort.py, fix old instructions.

objectivec/test/testdata/gen_models.sh

@@ -8,5 +8,5 @@ DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )"
 cd ${DIR}
 
 python3 ./single_add_gen.py
-python3 -m onnxruntime.tools.convert_onnx_models_to_ort --optimization_level basic .
 
+ORT_CONVERT_ONNX_MODELS_TO_ORT_OPTIMIZATION_LEVEL=basic python3 -m onnxruntime.tools.convert_onnx_models_to_ort --optimization_style=Fixed .

onnxruntime/test/optimizer/runtime_optimization/graph_runtime_optimization_test.cc

@@ -350,7 +350,7 @@ TEST(GraphRuntimeOptimizationTest, TestNhwcTransformerDirectlyUpdatesQLinearConv
       // - set environment variable ORT_CONVERT_ONNX_MODELS_TO_ORT_OPTIMIZATION_LEVEL=extended
       // - run:
       //     python -m onnxruntime.tools.convert_onnx_models_to_ort
-      //       --optimization_style Fixed
+      //       --optimization_style=Fixed
       //       testdata/transform/runtime_optimization/qdq_convs.onnx
       ORT_TSTR("testdata/transform/runtime_optimization/qdq_convs.extended.ort"),
       [](const OpCountMap& loaded_ops, const OpCountMap& initialized_ops) {

onnxruntime/test/testdata/mnist.readme.txt

@@ -5,6 +5,8 @@ We also save both ONNX and ORT format versions of the model with level 1 (aka 'b
     required_ops.config, which is used in the reduced ops CI build.
   - mnist.level1_opt.ort is used in NNAPI unit tests.
 
-The level 1 optimized model files can be generated by running the following command from the repo root and renaming the
-resulting .onnx and .ort files accordingly:
-  $ python ./tools/python/convert_onnx_models_to_ort.py --optimization_level basic --save_optimized_onnx_model ./onnxruntime/test/testdata/mnist.onnx
+The level 1 optimized model files can be generated with the following steps:
+- Set environment variable ORT_CONVERT_ONNX_MODELS_TO_ORT_OPTIMIZATION_LEVEL=basic
+- From this directory, run
+  $ python -m onnxruntime.tools.convert_onnx_models_to_ort --optimization_style=Fixed --save_optimized_onnx_model ./mnist.onnx
+- Rename the resulting .onnx and .ort files accordingly

onnxruntime/test/testdata/ort_backwards_compat/readme.md

@@ -1,13 +1,15 @@
 This directory contains ORT format models to test for backwards compatibility when we are forced to make an update that invalidates a kernel hash.
 
-When this happens, first create a new directory for the currently released ORT version.
+When this happens, first create a directory for the currently released ORT version if one doesn't already exist.
 
-Find a model that uses the operator with the kernel hash change. The ONNX test data is generally a good place to do this. See cmake/external/onnx/onnx/backend/test/data/node
+Find a model that uses the operator with the kernel hash change and copy it to the directory for the currently released ORT version.
+The ONNX test data is generally a good place to do this. See cmake/external/onnx/onnx/backend/test/data/node.
 
 Convert the model to ORT format using the currently released ORT version. This model will contain the original hash.
 
 e.g.
-Running `python -m onnxruntime.tools.convert_onnx_models_to_ort --optimization_level=basic ORTv1.10/not1.onnx` 
+Setting environment variable ORT_CONVERT_ONNX_MODELS_TO_ORT_OPTIMIZATION_LEVEL=basic
+and then running `python -m onnxruntime.tools.convert_onnx_models_to_ort --optimization_style=Fixed ORTv1.10/not1.onnx`
 will create the ORT format model `not1.basic.ort`
 
 Add both the ONNX and the ORT format models to the repository.

tools/python/util/convert_onnx_models_to_ort.py

@@ -189,11 +189,6 @@ def parse_args():
     parser.add_argument('--allow_conversion_failures', action='store_true',
                         help='Whether to proceed after encountering model conversion failures.')
 
-    parser.add_argument('--nnapi_partitioning_stop_ops',
-                        help='Specify the list of NNAPI EP partitioning stop ops. '
-                             'In particular, specify the value of the "ep.nnapi.partitioning_stop_ops" session '
-                             'options config entry.')
-
     parser.add_argument('--target_platform', type=str, default=None, choices=['arm', 'amd64'],
                         help='Specify the target platform where the exported model will be used. '
                              'This parameter can be used to choose between platform-specific options, '
@@ -226,9 +221,6 @@ def convert_onnx_models_to_ort():
 
     session_options_config_entries = {}
 
-    if args.nnapi_partitioning_stop_ops is not None:
-        session_options_config_entries["ep.nnapi.partitioning_stop_ops"] = args.nnapi_partitioning_stop_ops
-
     if args.target_platform == 'arm':
         session_options_config_entries["session.qdqisint8allowed"] = "1"
     else: