https://github.com/aistream-peelout/flow_docker
Files needed to run flow-forecast as a container.
https://github.com/aistream-peelout/flow_docker
deep-learning docker time-series-forecasting wandb
Last synced: about 1 month ago
JSON representation
Files needed to run flow-forecast as a container.
- Host: GitHub
- URL: https://github.com/aistream-peelout/flow_docker
- Owner: AIStream-Peelout
- Created: 2020-08-27T03:52:37.000Z (about 5 years ago)
- Default Branch: master
- Last Pushed: 2024-05-09T20:27:37.000Z (over 1 year ago)
- Last Synced: 2025-03-10T18:53:31.873Z (7 months ago)
- Topics: deep-learning, docker, time-series-forecasting, wandb
- Language: Python
- Homepage:
- Size: 59.6 KB
- Stars: 1
- Watchers: 2
- Forks: 1
- Open Issues: 6
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Flow Forecast Training Dockerfiles
This repository contains Dockerfiles needed to run Flow Forecast as a distributed Weights and Biases hyper-parameter sweep.
1. Make a sweep YAML file of the hyperparameters you want to search for on your FF model (example in the example_sweep directory). The program in the sweep file should be run_flow.py.
2. Get your WANDB_SWEEP_ID. For this you need to run `wandb sweep --project sweeps_demo config.yaml`
3. Make an environment file
In your environment file there should be the following items:
- WANDB_API_KEY
- WANDB_SWEEP_ID (this should be the full project/sweep)
- BASIC_CONFIG_PATH
- ENVIRONMENT_GCP (option)
1. Run `docker run --env-file .env_file aistream1/flow_sweep`
Your base config path file should have the JSON file with the default sweep values.