[GitHub] [tvm] mehrdadh commented on a change in pull request #9697: Add Module-Based Model Runtime Interface for AOT (support C++ runtime)

[GitBox <gi...@apache.org>]

mehrdadh commented on a change in pull request #9697:
URL: https://github.com/apache/tvm/pull/9697#discussion_r766992034



##########
File path: tests/python/relay/aot/test_cpp_aot.py
##########
@@ -0,0 +1,88 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements.  See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership.  The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License.  You may obtain a copy of the License at
+#
+#   http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied.  See the License for the
+# specific language governing permissions and limitations
+# under the License.
+
+
+import sys
+import textwrap
+
+import numpy as np
+import pytest
+
+import tvm
+from tvm import relay, TVMError
+from tvm.ir.module import IRModule
+from tvm.relay import testing, transform
+from tvm.relay.testing import byoc
+from tvm.relay.op.annotation import compiler_begin, compiler_end
+from aot_test_utils import (
+    AOTTestModel,
+    AOT_DEFAULT_RUNNER,
+    generate_ref_data,
+    convert_to_relay,
+    compile_and_run,
+    compile_models,
+    parametrize_aot_options,
+)
+
+
+def test_conv2d():
+    RELAY_MODEL = textwrap.dedent(

Review comment:
       I suggest to use relay directly to build the relay model. Here is an example:
   https://github.com/apache/tvm/blob/01599d13777b0a8723c6a2c940afa95fd863d5ee/tests/micro/zephyr/test_zephyr.py#L394

##########
File path: CMakeLists.txt
##########
@@ -390,6 +391,13 @@ if(USE_PROFILER)
   list(APPEND RUNTIME_SRCS ${RUNTIME_VM_PROFILER_SRCS})
 endif(USE_PROFILER)
 
+if(USE_AOT_EXECUTOR)

Review comment:
       I suggest to add `USE_AOT_EXECUTOR` to cmake/config.cmake with default value ON

##########
File path: tests/python/relay/aot/test_cpp_aot.py
##########
@@ -0,0 +1,88 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements.  See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership.  The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License.  You may obtain a copy of the License at
+#
+#   http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied.  See the License for the
+# specific language governing permissions and limitations
+# under the License.
+
+
+import sys
+import textwrap
+
+import numpy as np
+import pytest
+
+import tvm
+from tvm import relay, TVMError
+from tvm.ir.module import IRModule
+from tvm.relay import testing, transform
+from tvm.relay.testing import byoc
+from tvm.relay.op.annotation import compiler_begin, compiler_end
+from aot_test_utils import (
+    AOTTestModel,
+    AOT_DEFAULT_RUNNER,
+    generate_ref_data,
+    convert_to_relay,
+    compile_and_run,
+    compile_models,
+    parametrize_aot_options,
+)
+
+
+def test_conv2d():
+    RELAY_MODEL = textwrap.dedent(
+        """\
+        #[version = "0.0.5"]
+        def @main(%data : Tensor[(1, 3, 64, 64), uint8], %weight : Tensor[(8, 3, 5, 5), int8]) {
+            %1 = nn.conv2d(
+                 %data,
+                 %weight,
+                 padding=[2, 2],
+                 channels=8,
+                 kernel_size=[5, 5],
+                 data_layout="NCHW",
+                 kernel_layout="OIHW",
+                 out_dtype="int32");
+          %1
+        }
+    """
+    )
+    ir_mod = tvm.parser.fromtext(RELAY_MODEL)
+
+    main_func = ir_mod["main"]
+    shape_dict = {p.name_hint: p.checked_type.concrete_shape for p in main_func.params}
+    type_dict = {p.name_hint: p.checked_type.dtype for p in main_func.params}
+
+    weight_data = np.ones(shape_dict["weight"]).astype(type_dict["weight"])
+    input_data = np.ones(shape_dict["data"]).astype(type_dict["data"])
+
+    params = {"weight": weight_data}
+    inputs = {"data": input_data}
+    output_list = generate_ref_data(ir_mod, inputs, params)
+
+    with tvm.transform.PassContext(opt_level=3, config={"tir.disable_vectorize": True}):
+        mod = tvm.relay.build(ir_mod, params=params, target="c -executor=aot -link-params")

Review comment:
       change this to use `executor` and `runtime` args?




-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org