...
Below is the list of Python libraries in the ds and ds-cuda9-cudnn7 tags.
Library | Version |
|---|---|
simplejson | 3.16.0 |
requests | 2.20.0 |
jsonpickle | 1.0 |
python-dateutil | 2.7.4 |
more-itertools | 4.3.0 |
pydub | 0.23.0 |
numpy | 1.15.3 |
scipy | 1.1.0 |
scikit-learn | 0.20.0 |
xgboost | 0.80 |
lightgbm | 2.2.1 |
pillow | 5.3.0 |
bokeh | 0.13.0 |
pandas | 0.23.4 |
tensorflow | 1.5.0 |
keras | 2.2.4 |
nltk | 3.3 |
textblob | 0.15.1 |
| Note |
|---|
ds-cuda9-cudnn7 has tensorflow-gpu==1.11.0 instead of tensorflow==1.5.0. |
...
The following table describes each option:
Option | Format | Description |
|---|---|---|
--memory-swap="-1" --restart=always | --memory-swap="<memory_swap_limit>" --restart=always | The container can use unlimited space for memory swapping. See Memory Swap for more information. "–restart=always" automatically restarts the docker container if the machine restarts. |
-p 5301:5301 | -p <host_port>:5301 | The RPE is accessible from the host port which is 5301 by default. You may change the host port to support multiple containers on the same node. |
-e "REMOTE_PYTHON_EXECUTOR_TOKEN=" | -e "REMOTE_PYTHON_EXECUTOR_TOKEN=<token>" | The default token is empty. A strong token is recommended. |
-v /opt/remote_python_executor_log/:/opt/remote_python_executor_log/ | -v <log_dir>/opt/remote_python_executor_log/ | The log is mounted to /opt/remote_python_executor_log/by default. This location can be changed. |
--name=rpe | --name=<container_name> | The container’s name can be changed in case of multiple containers on the same node. |
Manage RPE
Since the RPE starts with the container, use the following commands to start, stop, restart and remove the RPE.
| Code Block |
|---|
sudo docker start rpe
sudo docker stop rpe
sudo docker restart rpe
sudo docker rm rpe |
Custom Image
Download the custom RPE package. This package contains Dockerfile and others. You can modify the Dockerfile and add required libraries in requirements.txt. Then, run the following command to build the Docker image.
...