Skip to content

Latest commit

 

History

History
97 lines (64 loc) · 3.11 KB

keras2cntk.md

File metadata and controls

97 lines (64 loc) · 3.11 KB

Keras "inception_v3" to CNTK conversion example

Model: "inception_v3" for ImageNet

Source: Keras 2.1.3

Destination: CNTK 2.4

Framework Installation

Install Keras and CNTK in case

$ pip install keras

$ pip install https://cntk.ai/PythonWheel/CPU-Only/cntk-2.4-cp27-cp27mu-linux_x86_64.whl
or
$ pip install https://cntk.ai/PythonWheel/CPU-Only/cntk-2.4-cp35-cp35m-linux_x86_64.whl

Keras Model Preparation

First, you'll need to prepare your pre-trained keras model. There is a pre-trained model extractor for frameworks to do this. Refer to it to extract your Keras model's structure and weights.

$ mmdownload -f keras -n inception_v3

Keras model inception_v3 is saved in [./imagenet_inception_v3.h5]

Then you got the Keras pre-trained inception_v3 model which is downloaded to the current working directory.

Convert Keras Model to CNTK

We provide two ways to convert models.

One-step Command

For versions above [email protected], we provide one command to achieve converting a Keras Model to CNTK.

$ mmconvert -sf keras -iw imagenet_inception_v3.h5 -df cntk -om keras_to_cntk_inception_v3.dnn
.
.
.
CNTK model file is saved as [keras_to_cntk_inception_v3.dnn], generated by [2c33f7f278cb46be992f50226fcfdb5d.py] and [2c33f7f278cb46be992f50226fcfdb5d.npy].

Now you've got the CNTK original model keras_to_cntk_inception_v3.dnn, converted from Keras in one step!. 2c33f7f278cb46be992f50226fcfdb5d.py and 2c33f7f278cb46be992f50226fcfdb5d.npy are temporal files which will be removed automatically.

Step-by-step Command (for debugging)

Convert the pre-trained model files to intermediate representation

$ mmtoir -f keras -w imagenet_inception_v3.h5 -o converted

Using TensorFlow backend.
.
.
.
IR network structure is saved as [converted.json].
IR network structure is saved as [converted.pb].
IR weights are saved as [converted.npy].

Then you got the intermediate representation files converted.json for visualization, converted.pb and converted.npy for next steps.

Convert the IR files to CNTK models

$ mmtocode -f cntk -d converted_cntk.py -n converted.pb -w converted.npy

Parse file [converted.pb] with binary format successfully.
Target network code snippet is saved as [converted_cntk.py].

And you will get a file named converted_cntk.py, which contains the original CNTK codes to build the Inception V3 network.

With these three steps, you have already converted the pre-trained Keras Inception_v3 models to CNTK network file converted_cntk.py and weight file converted.npy. You can use these two files to fine-tune training or inference.

Dump the original CNTK model

$ mmtomodel -f cntk -in converted_cntk -iw converted.npy -o cntk_inception_v3.dnn
.
.
.
CNTK model file is saved as [cntk_inception_v3.dnn], generated by [converted_cntk.py] and [converted.npy].

The file cntk_inception_v3.dnn can be loaded by CNTK directly.