Thanks to visit codestin.com
Credit goes to github.com

Skip to content

Commit 7b549ae

Browse files
Roshrinilupesko
Roshrini
authored and
lupesko
committed
Update mxnet doc and adding MXNet to ONNX export tutorial (#50)
* mxnet export tutorial added * doc fix * small fixes * adding details for supported onnx operator opset
1 parent 0f308c2 commit 7b549ae

3 files changed

Lines changed: 277 additions & 6 deletions

File tree

‎README.md‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -7,7 +7,7 @@
77
| [Caffe2](http://caffe2.ai) | [part of caffe2 package](https://github.com/caffe2/caffe2/tree/master/caffe2/python/onnx) | [Exporting](tutorials/Caffe2OnnxExport.ipynb) | [Importing](tutorials/OnnxCaffe2Import.ipynb) |
88
| [PyTorch](http://pytorch.org/) | [part of pytorch package](http://pytorch.org/docs/master/onnx.html) | [Exporting](tutorials/PytorchOnnxExport.ipynb), [Extending support](tutorials/PytorchAddExportSupport.md) | coming soon |
99
| [Cognitive Toolkit (CNTK)](https://www.microsoft.com/en-us/cognitive-toolkit/) | [built-in](https://docs.microsoft.com/en-us/cognitive-toolkit/setup-cntk-on-your-machine) | [Exporting](tutorials/CntkOnnxExport.ipynb) | [Importing](tutorials/OnnxCntkImport.ipynb) |
10-
| [Apache MXNet](http://mxnet.incubator.apache.org/) | part of mxnet package [docs](http://mxnet.incubator.apache.org/versions/1.2.0/api/python/contrib/onnx.html) [github](https://github.com/apache/incubator-mxnet/tree/master/python/mxnet/contrib/onnx) | coming soon | [Importing](tutorials/OnnxMxnetImport.ipynb) [experimental] |
10+
| [Apache MXNet](http://mxnet.incubator.apache.org/) | part of mxnet package [docs](http://mxnet.incubator.apache.org/api/python/contrib/onnx.html) [github](https://github.com/apache/incubator-mxnet/tree/master/python/mxnet/contrib/onnx) | [Exporting](tutorials/MXNetONNXExport.ipynb) | [Importing](tutorials/OnnxMxnetImport.ipynb) |
1111
| [Chainer](https://chainer.org/) | [chainer/onnx-chainer](https://github.com/chainer/onnx-chainer) | [Exporting](tutorials/ChainerOnnxExport.ipynb) | coming soon |
1212
| [TensorFlow](https://www.tensorflow.org/) | [onnx/onnx-tensorflow](https://github.com/onnx/onnx-tensorflow) | [Exporting](tutorials/OnnxTensorflowExport.ipynb) | [Importing](tutorials/OnnxTensorflowImport.ipynb) [experimental] |
1313
| [Apple CoreML](https://developer.apple.com/documentation/coreml) | [onnx/onnx-coreml](https://github.com/onnx/onnx-coreml) and [onnx/onnxmltools](https://github.com/onnx/onnxmltools) | [Exporting](https://github.com/onnx/onnxmltools) | [Importing](tutorials/OnnxCoremlImport.ipynb) |

‎tutorials/MXNetONNXExport.ipynb‎

Lines changed: 265 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,265 @@
1+
{
2+
"cells": [
3+
{
4+
"cell_type": "markdown",
5+
"metadata": {},
6+
"source": [
7+
"# Exporting MXNet models to ONNX\n",
8+
"\n",
9+
"In this tutorial, we will show how you can save MXNet models to ONNX format.\n",
10+
"ONNX exporter it a part of [MXNet repository](https://github.com/apache/incubator-mxnet/tree/master/python/mxnet/contrib/onnx/mx2onnx).\n",
11+
"\n",
12+
"Current MXNet-ONNX import and export operator support and coverage can be found [here](Operator support and coverage - https://cwiki.apache.org/confluence/display/MXNET/ONNX):\n"
13+
]
14+
},
15+
{
16+
"cell_type": "markdown",
17+
"metadata": {},
18+
"source": [
19+
"## Step 1: Installations:\n",
20+
"#### Install **ONNX 1.2.1 version.** Follow instructions on [ONNX repo](https://github.com/onnx/onnx).\n",
21+
"\n",
22+
"MXNet ONNX importer and exporter follows version 7 of ONNX operator set.\n",
23+
"\n",
24+
"#### Make sure to install latest MXNet either from source or pip.\n",
25+
"\n",
26+
"```bash\n",
27+
"pip install mxnet --pre\n",
28+
"```\n",
29+
"\n",
30+
"* Note: ONNX exporter will be released as a part of MXNet v1.3. "
31+
]
32+
},
33+
{
34+
"cell_type": "markdown",
35+
"metadata": {},
36+
"source": [
37+
"## Step 2: Prepare MXNet model to convert to ONNX:\n",
38+
"\n",
39+
"Let's try out pretrained resnet model from [MXNet model zoo](http://data.mxnet.io/models/)."
40+
]
41+
},
42+
{
43+
"cell_type": "code",
44+
"execution_count": 1,
45+
"metadata": {
46+
"collapsed": true
47+
},
48+
"outputs": [],
49+
"source": [
50+
"import mxnet as mx\n",
51+
"import numpy as np\n",
52+
"from mxnet.contrib import onnx as onnx_mxnet"
53+
]
54+
},
55+
{
56+
"cell_type": "code",
57+
"execution_count": 2,
58+
"metadata": {},
59+
"outputs": [
60+
{
61+
"data": {
62+
"text/plain": [
63+
"['resnet-18-0000.params', 'resnet-18-symbol.json', 'synset.txt']"
64+
]
65+
},
66+
"execution_count": 2,
67+
"metadata": {},
68+
"output_type": "execute_result"
69+
}
70+
],
71+
"source": [
72+
"# Download pretrained resnet model - json and params from mxnet model zoo.\n",
73+
"path='http://data.mxnet.io/models/imagenet/'\n",
74+
"[mx.test_utils.download(path+'resnet/18-layers/resnet-18-0000.params'),\n",
75+
" mx.test_utils.download(path+'resnet/18-layers/resnet-18-symbol.json'),\n",
76+
" mx.test_utils.download(path+'synset.txt')]"
77+
]
78+
},
79+
{
80+
"cell_type": "markdown",
81+
"metadata": {},
82+
"source": [
83+
"## Step 3: Use MXNet to ONNX exporter:\n",
84+
"\n",
85+
"MXNet's ONNX \"export_model\" API accepts following inputs: \n"
86+
]
87+
},
88+
{
89+
"cell_type": "code",
90+
"execution_count": 3,
91+
"metadata": {},
92+
"outputs": [
93+
{
94+
"name": "stdout",
95+
"output_type": "stream",
96+
"text": [
97+
"Help on function export_model in module mxnet.contrib.onnx.mx2onnx.export_model:\n",
98+
"\n",
99+
"export_model(sym, params, input_shape, input_type=<type 'numpy.float32'>, onnx_file_path=u'model.onnx', verbose=False)\n",
100+
" Exports the MXNet model file, passed as a parameter, into ONNX model.\n",
101+
" Accepts both symbol,parameter objects as well as json and params filepaths as input.\n",
102+
" Operator support and coverage - https://cwiki.apache.org/confluence/display/MXNET/ONNX\n",
103+
" \n",
104+
" Parameters\n",
105+
" ----------\n",
106+
" sym : str or symbol object\n",
107+
" Path to the json file or Symbol object\n",
108+
" params : str or symbol object\n",
109+
" Path to the params file or params dictionary. (Including both arg_params and aux_params)\n",
110+
" input_shape : List of tuple\n",
111+
" Input shape of the model e.g [(1,3,224,224)]\n",
112+
" input_type : data type\n",
113+
" Input data type e.g. np.float32\n",
114+
" onnx_file_path : str\n",
115+
" Path where to save the generated onnx file\n",
116+
" verbose : Boolean\n",
117+
" If true will print logs of the model conversion\n",
118+
" \n",
119+
" Returns\n",
120+
" -------\n",
121+
" onnx_file_path : str\n",
122+
" Onnx file path\n",
123+
"\n"
124+
]
125+
}
126+
],
127+
"source": [
128+
"help(onnx_mxnet.export_model)"
129+
]
130+
},
131+
{
132+
"cell_type": "markdown",
133+
"metadata": {},
134+
"source": [
135+
"From the API description, you can see that `export_model` API accepts 2 kinds of inputs:\n",
136+
"\n",
137+
"### MXNet sym, params objects:\n",
138+
"\n",
139+
" This is useful if you are training a model. At the end of training, you just need to invoke the `export_model` function and provide sym and params objects as inputs with other attributes to save the model in ONNX format.\n",
140+
" \n",
141+
"### MXNet's exported json and params files:\n",
142+
"\n",
143+
" This is useful if you have pretrained models and you want to convert them to ONNX format.\n",
144+
"\n",
145+
"In this tutorial, we will show second usecase: "
146+
]
147+
},
148+
{
149+
"cell_type": "code",
150+
"execution_count": 4,
151+
"metadata": {
152+
"collapsed": true
153+
},
154+
"outputs": [],
155+
"source": [
156+
"# Downloaded input symbol and params files\n",
157+
"sym = 'resnet-18-symbol.json'\n",
158+
"params = 'resnet-18-0000.params'\n",
159+
"# Standard Imagenet input - 3 channels, 224*224\n",
160+
"input_shape = (1,3,224,224)\n",
161+
"# Path of the output file\n",
162+
"onnx_file = 'mxnet_exported_resnet50.onnx'"
163+
]
164+
},
165+
{
166+
"cell_type": "code",
167+
"execution_count": 5,
168+
"metadata": {
169+
"collapsed": true
170+
},
171+
"outputs": [],
172+
"source": [
173+
"# Invoke export model API. It returns path of the converted onnx model\n",
174+
"converted_model_path = onnx_mxnet.export_model(sym, params, [input_shape], np.float32, onnx_file)"
175+
]
176+
},
177+
{
178+
"cell_type": "code",
179+
"execution_count": 6,
180+
"metadata": {},
181+
"outputs": [
182+
{
183+
"name": "stdout",
184+
"output_type": "stream",
185+
"text": [
186+
"mxnet_exported_resnet50.onnx\n"
187+
]
188+
}
189+
],
190+
"source": [
191+
"print(converted_model_path)"
192+
]
193+
},
194+
{
195+
"cell_type": "markdown",
196+
"metadata": {},
197+
"source": [
198+
"## Step 4: Check validity\n",
199+
"\n",
200+
"You can check validity of the converted ONNX model by using ONNX checker tool as follows:"
201+
]
202+
},
203+
{
204+
"cell_type": "code",
205+
"execution_count": 7,
206+
"metadata": {},
207+
"outputs": [],
208+
"source": [
209+
"from onnx import checker\n",
210+
"import onnx\n",
211+
"# Load onnx model\n",
212+
"model_proto = onnx.load(converted_model_path)\n",
213+
"\n",
214+
"# Check if converted ONNX protobuf is valid\n",
215+
"checker.check_graph(model_proto.graph)"
216+
]
217+
},
218+
{
219+
"cell_type": "markdown",
220+
"metadata": {},
221+
"source": [
222+
"This method confirms input model protobuf is valid."
223+
]
224+
},
225+
{
226+
"cell_type": "markdown",
227+
"metadata": {},
228+
"source": [
229+
"## What's next\n",
230+
"\n",
231+
"Take a look at [other tutorials, including importing of ONNX models to other frameworks](https://github.com/onnx/tutorials/tree/master/tutorials)"
232+
]
233+
},
234+
{
235+
"cell_type": "code",
236+
"execution_count": null,
237+
"metadata": {
238+
"collapsed": true
239+
},
240+
"outputs": [],
241+
"source": []
242+
}
243+
],
244+
"metadata": {
245+
"kernelspec": {
246+
"display_name": "onnx_mxnet",
247+
"language": "python",
248+
"name": "onnx_mxnet"
249+
},
250+
"language_info": {
251+
"codemirror_mode": {
252+
"name": "ipython",
253+
"version": 2
254+
},
255+
"file_extension": ".py",
256+
"mimetype": "text/x-python",
257+
"name": "python",
258+
"nbconvert_exporter": "python",
259+
"pygments_lexer": "ipython2",
260+
"version": "2.7.14"
261+
}
262+
},
263+
"nbformat": 4,
264+
"nbformat_minor": 2
265+
}

‎tutorials/OnnxMxnetImport.ipynb‎

Lines changed: 11 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -6,7 +6,7 @@
66
"source": [
77
"# Importing ONNX models to MXNet\n",
88
"\n",
9-
"We'll show how you can use ONNX-MXNet to import ONNX models into MXNet, and use the imported model for inference, benefiting from MXNet's optimized execution engine.\n",
9+
"We'll show how you can import ONNX models into MXNet, and use the imported model for inference, benefiting from MXNet's optimized execution engine.\n",
1010
"\t\n",
1111
"## Step 1: Installations\n",
1212
"\t\n",
@@ -31,13 +31,15 @@
3131
"\n",
3232
"## Step 3: Import the ONNX model into MXNet\n",
3333
"\n",
34-
"Now that we have an ONNX model file ready, let's import it into MXNet using ONNX-MXNet's import API."
34+
"Now that we have an ONNX model file ready, let's import it into MXNet:"
3535
]
3636
},
3737
{
3838
"cell_type": "code",
3939
"execution_count": 12,
40-
"metadata": {},
40+
"metadata": {
41+
"collapsed": true
42+
},
4143
"outputs": [],
4244
"source": [
4345
"import mxnet.contrib.onnx as onnx_mxnet\n",
@@ -95,7 +97,9 @@
9597
{
9698
"cell_type": "code",
9799
"execution_count": 14,
98-
"metadata": {},
100+
"metadata": {
101+
"collapsed": true
102+
},
99103
"outputs": [],
100104
"source": [
101105
"mod = mx.mod.Module(symbol=sym, data_names=['1'], label_names=None)\n",
@@ -114,7 +118,9 @@
114118
{
115119
"cell_type": "code",
116120
"execution_count": 15,
117-
"metadata": {},
121+
"metadata": {
122+
"collapsed": true
123+
},
118124
"outputs": [],
119125
"source": [
120126
"from collections import namedtuple\n",

0 commit comments

Comments
 (0)