Skip to main content
Version: Next

Output/Working directory

Β Example (Click Here)

Hydra can solve the problem of your needing to specify a new output directory for each run, by creating a directory for each run and executing your code within that output directory. By default, this output directory is used to store Hydra output for the run (Configuration, Logs etc).

Every time you run the app, a new output directory is created. You can retrieve the path of the output directy by inspecting the Hydra config as in the example below.

my_app.py
import os
from omegaconf import DictConfig
import hydra

@hydra.main(version_base=None)
def my_app(_cfg: DictConfig) -> None:
print(f"Working directory : {os.getcwd()}")
print(f"Output directory : {hydra.core.hydra_config.HydraConfig.get().runtime.output_dir}")
$ python my_app.py
Working directory : /home/omry/dev/hydra
Output directory : /home/omry/dev/hydra/outputs/2019-09-25/15-16-17

$ python my_app.py
Working directory : /home/omry/dev/hydra
Output directory : /home/omry/dev/hydra/outputs/2019-09-25/15-16-19

Let's take a look at one of the output directories:

$ tree outputs/2019-09-25/15-16-17
outputs/2019-09-25/15-16-17
β”œβ”€β”€ .hydra
β”‚ β”œβ”€β”€ config.yaml
β”‚ β”œβ”€β”€ hydra.yaml
β”‚ └── overrides.yaml
└── my_app.log

We have the Hydra output directory (.hydra by default), and the application log file. Inside the Hydra output directory we have:

  • config.yaml: A dump of the user specified configuration
  • hydra.yaml: A dump of the Hydra configuration
  • overrides.yaml: The command line overrides used

And in the main output directory:

  • my_app.log: A log file created for this run

You can configure the name of the output directory using the customizing the working directory pattern.

Automatically change current working dir to job's output dir​

By setting hydra.job.chdir=True, you can configure Hydra's @hydra.main decorator to change python's working directory by calling os.chdir before passing control to the user's decorated main function. As of Hydra v1.2, hydra.job.chdir defaults to False. Setting hydra.job.chdir=True enables convenient use of the output directory to store output for the application (For example, a database dump file).

# check current working dir
$ pwd
/home/jasha/dev/hydra

# for Hydra >= 1.2, working dir remains unchanged by default
$ python my_app.py
Working directory : /home/jasha/dev/hydra
Output directory : /home/jasha/dev/hydra/outputs/2023-04-18/13-43-24

# working dir changed to output dir
$ python my_app.py hydra.job.chdir=True
Working directory : /home/jasha/dev/hydra/outputs/2023-04-18/13-43-17
Output directory : /home/jasha/dev/hydra/outputs/2023-04-18/13-43-17

# output dir and files are still created even if `chdir` is disabled:
$ tree -a outputs/2023-04-18/13-43-24/
outputs/2023-04-18/13-43-24/
β”œβ”€β”€ .hydra
β”‚Β Β  β”œβ”€β”€ config.yaml
β”‚Β Β  β”œβ”€β”€ hydra.yaml
β”‚Β Β  └── overrides.yaml
└── my_app.log

Changing or disabling Hydra's output subdir​

You can change the .hydra subdirectory name by overriding hydra.output_subdir. You can disable its creation by overriding hydra.output_subdir to null.

Accessing the original working directory in your application​

With hydra.job.chdir=True, you can still access the original working directory by importing get_original_cwd() and to_absolute_path() in hydra.utils:

from hydra.utils import get_original_cwd, to_absolute_path

@hydra.main(version_base=None)
def my_app(_cfg: DictConfig) -> None:
print(f"Current working directory : {os.getcwd()}")
print(f"Orig working directory : {get_original_cwd()}")
print(f"to_absolute_path('foo') : {to_absolute_path('foo')}")
print(f"to_absolute_path('/foo') : {to_absolute_path('/foo')}")

if __name__ == "__main__":
my_app()
$ python examples/tutorial/8_working_directory/original_cwd.py
Current working directory  : /Users/omry/dev/hydra/outputs/2019-10-23/10-53-03
Original working directory : /Users/omry/dev/hydra
to_absolute_path('foo') : /Users/omry/dev/hydra/foo
to_absolute_path('/foo') : /foo

The name of the generated working directories can be customized.