[GitHub] [incubator-tvm] zhiics commented on a change in pull request #6355: [BYOC][ETHOSN] Introduce further operator support

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

zhiics commented on a change in pull request #6355:
URL: https://github.com/apache/incubator-tvm/pull/6355#discussion_r485769865



##########
File path: src/relay/backend/contrib/ethosn/codegen.cc
##########
@@ -304,6 +475,42 @@ EthosnError ConstructNetworkVisitor::MakeSplitLayer(const Call& call, sl::Tensor
   return EthosnError();
 }
 
+EthosnError ConstructNetworkVisitor::MakeDepthToSpaceLayer(const Call& call,
+                                                           sl::TensorAndId<sl::Operand>* out) {
+  DepthToSpaceParams params;
+  params.input_info = GetTensorInfo(tensor_table_, call);
+  if (auto err = EthosnAPI::DepthToSpace(call, &params)) {
+    return err;
+  }
+
+  auto input = operand_table_[call->args[0]][0];
+
+  try {
+    *out = AddDepthToSpace(network_, *input, params.depth_info);
+  } catch (const sl::NotSupportedException& e) {
+    return EthosnError(e.what());
+  }
+  return EthosnError();
+}
+
+EthosnError ConstructNetworkVisitor::MakeReluLayer(const Call& call,

Review comment:
       Just a minor suggestion that you feel free to ignore, I think these functions can be replaced by a macro as they look exactly the same and only operator name matters. 

##########
File path: tests/python/contrib/test_ethosn/test_networks.py
##########
@@ -0,0 +1,163 @@
+# 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.
+
+"""Ethos-N integration end-to-end network tests"""
+
+import pytest
+pytest.importorskip('tflite')
+pytest.importorskip('tensorflow')
+
+from tvm import relay
+from tvm.relay.op.contrib.ethosn import ethosn_available, Available
+from tvm.contrib import download
+import tvm.relay.testing.tf as tf_testing
+import tflite.Model
+from . import infrastructure as tei
+
+
+def _get_tflite_model(tflite_model_path, inputs_dict, dtype):
+    with open(tflite_model_path, 'rb') as f:
+        tflite_model_buffer = f.read()
+
+    try:
+        tflite_model = tflite.Model.Model.GetRootAsModel(tflite_model_buffer, 0)
+    except AttributeError:
+        tflite_model = tflite.Model.GetRootAsModel(tflite_model_buffer, 0)
+    shape_dict = {}
+    dtype_dict = {}
+    for input in inputs_dict:
+        input_shape = inputs_dict[input]
+        shape_dict[input] = input_shape
+        dtype_dict[input] = dtype
+
+    return relay.frontend.from_tflite(
+        tflite_model,
+        shape_dict=shape_dict,
+        dtype_dict=dtype_dict,
+    )
+
+
+def _test_image_network(model_url, model_sub_path, input_dict, compile_hash, output_count, run=True, host_ops=0, npu_partitions=1):
+    if not ethosn_available():
+        return
+
+    def get_model():
+        if model_url[-3:] in ("tgz", "zip"):
+            model_path = tf_testing.get_workload_official(
+                model_url,
+                model_sub_path,
+            )
+        else:
+            model_path = download.download_testdata(
+                model_url,
+                model_sub_path,
+            )
+        return _get_tflite_model(model_path, input_dict, 'uint8')
+
+    outputs = []
+    inputs = {}
+    for input_name in input_dict:
+        input_shape = input_dict[input_name]
+        inputs[input_name] = tei.get_real_image(input_shape[1], input_shape[2])
+
+    for npu in [False, True]:
+        mod, params = get_model()
+        graph, lib, params = tei.build(mod, params, npu=npu, expected_host_ops=host_ops, npu_partitions=npu_partitions)
+        if npu:
+            tei.assert_lib_hash(lib, compile_hash)

Review comment:
       what the purpose to compare a compiled lib to a predefined hash?  I understand constructing a full json graph is impractical when the program is large, but I think you can probably compare the some nodes in the json graph or assert the number of the nodes you would have in the graph and some representative ones in it.




----------------------------------------------------------------
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.

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