Custom Trial Environments
A trial runner is a Docker container that provides an isolated environment for running deep learning workloads. A trial runner consists of a set of Python libraries and some PEDL-specific harness code that orchestrates running workloads inside the container. When the container is launched, PEDL injects the current experiment's model definition, along with the values of the hyperparameters for the current trial.
In the current version of PEDL, all experiments use one of two base container images:
determinedai/pedl-tr-py3.6-tf, which includes Python 3.6.8, TensorFlow 1.12.0, Keras 2.2.4, and NumPy 1.15.0. This image is selected by default if the experiment configuration does not specify
determinedai/pedl-tr-py3.6-pytorch, which includes Python 3.6.8, PyTorch 0.4.1, torchvision 0.2.1, and NumPy 1.15.0.
PEDL will automatically select a GPU-enabled version of the image when running on agents with GPUs. The trial environment can be further customized by running an arbitrary list of "runtime" commands and installing additional Python packages, as described below.
Custom Trial Environments¶
If the model definition has publicly accessible dependencies that are not included in the base trial runner image, the experiment config file provides two options for adding to the trial runner environment:
runtime_commands. Before running the experiment, PEDL will generate and cache a new trial runner container with the requested packages installed and commands executed.
Here is an example experiment configuration that uses the
trial_environment: base_image: determinedai/pedl-tr-py3.6-tf runtime_packages: - pandas - numpy
This instructs PEDL to install these packages (via
pip) into the trial container before running any workloads.
Packages can additionally be restricted with version specifiers or installed from custom external sources using the standard
pip install syntax:
trial_environment: base_image: determinedai/pedl-tr-py3.6-tf runtime_packages: - pandas==0.20.3 # Install this exact version - pandas>=0.20.0 # Install a version greater than or equal to this version - pandas!=0.20.1 # Exclude this version from installation - git+https://github.com/pandas-dev/pandas # Installing from GitHub
runtime_packages makes it easy to customize the trial environment to include additional Python packages. For more complex changes to the trial environment (e.g., installing native libraries), users can specify the
trial_environment: base_image: determinedai/pedl-tr-py3.6-tf runtime_commands: - echo "Installing python3-pandas via apt-get" - apt-get install python3-pandas
The contents of
runtime_commands will be executed in the order provided and before any
runtime_packages are installed.
runtime_packages support installing different dependencies for GPU vs. CPU environments. For example:
trial_environment: base_image: determinedai/pedl-tr-py3.6-tf runtime_packages: cpu: - tensorflow gpu: - tensorflow-gpu
PEDL also provides the option to install dependencies from a user's local environment, as long as the local package is a Python source distribution file, or a Python wheel file that supports installation into a
py3 environment. To use a local package dependency, specify the
--package flag when creating the experiment with
pedl experiment create:
$ pedl experiment create config.yaml model_def.py \ --package user-package-1.tar.gz \ --package user-package-2.tar.gz
The command above will install
user-package-2.tar.gz on all trial runners in the experiment. There is no limit on the number of packages that can be specified. However the total size of the model definition and all packages must be under 96 MB.
Trial Runner Environment¶
PEDL allows users to configure the environment used by trial runner containers. If the model definition directory contains a file named
pedl-prepare-env.sh, this file will be executed (via
bash) before the trial runner's Python environment is started. This feature can be used to set environmental variables or run commands that prepare the trial environment.
Custom Docker Base Images¶
This is currently an experimental feature.
PEDL supports running trial containers that are based on user-provided Docker images. The
base_image should be accessible to all agent nodes via
docker pull. If a private image is used, Docker Registry credentials must be specified in the
registry_auth section in the experiment configuration.
When provided with a custom base image, PEDL will build a runtime image that imports the
base_image with a
FROM instruction, injects the Python PEDL source code with
COPY instructions, and installs Python requirement libraries with
pip. The maintainer of the custom base image is responsible for installing the following PEDL dependencies:
- Git >= 1.5.0
- Python 3.6.8 installed as
- CUDA 9.0
- CuDNN 7.4
- Framework-specific libraries, depending on the interface used:
- TensorFlow 1.12.0 if using
- PyTorch 0.4.1 if using the
- TensorFlow 1.12.0 if using
If you would like to use versions of libraries different from those specified above in a custom base image, please contact the Determined AI team for a consultation.