Use with Chainer model
======================

In this tutorial, we'll convert ResNet50 [#f1]_ classification model pretrained in Chainer [#f2]_ into WebDNN execution format.

1. Load chainer pretrained model

.. code-block:: python

    import chainer

    model = chainer.links.model.vision.resnet.ResNet50Layers()

2. Execute model with dummy data. In chainer, computation graph are defined by run. Therefore we need execute model at least once to construct the graph.

.. code-block:: python

    import numpy as np

    x = chainer.Variable(np.empty((1, 3, 224, 224), dtype=np.float32))
    y = model(x, layers=["fc6"])["fc6"]

3. Convert chainer computation graph to our computation graph format

.. code-block:: python

    from webdnn.frontend.chainer import ChainerConverter

    graph = ChainerConverter().convert([x], [y])

4. Generate and save execution information.

.. code-block:: python

    from webdnn.backend import generate_descriptor

    exec_info = generate_descriptor("webgpu", graph)  # also "webassembly", "webgl", "fallback" are available.
    exec_info.save("./output")

To run converted model on web browser, please see :ref:`"#3. Run on web browser" in keras tutorial<js-api>`

.. rubric:: References
.. [#f1] K. He, X. Zhang, S. Ren, and J. Sun, "Deep Residual Learning for Image Recognition", IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016.
.. [#f2] http://docs.chainer.org/en/latest/reference/links.html#chainer.links.ResNet50Layers