Closed p16i closed 5 years ago
Hi @kittinan and @chanwit, can someone review this PR?
Deepcut using default Keras configuration. So we don't have to add the config file to Docker image.
Deepcut Keras config:
{
"floatx": "float32",
"epsilon": 1e-07,
"backend": "tensorflow",
"image_data_format": "channels_last"
}
Keras default config: https://keras.io/backend/#switching-from-one-backend-to-another
This pull request will fix Issue #36
LGTM
Seems like I can merge the PR then!
In case of having a custom Keras file, one could bind a local Keras file into the container like:
docker run -v $PWD/keras.json:/root/.keras/keras.json deepcut
Thanks!
Actually, it would be more convenient if we have a built DeepCut image hosted on hub.docker.com
.
Please someone setup the deepcut organization there on Docker hub, and grant me an access. I'll help maintain the images.
Basically, the image name would be something like deepcut/deepcut:latest
Hi,
It seems
keras.json
[1] isn't in the repository anymore. However, it's still inDockerfile
and causes docker-build failed.[1] https://github.com/rkcosmos/deepcut/blob/master/Dockerfile#L39