Configure hyperparameters from the CLI (Advanced)¶

Customize arguments by subcommand¶

To customize arguments by subcommand, pass the config before the subcommand:

$ python main.py [before] [subcommand] [after]
$ python main.py  ...         fit       ...

For example, here we set the Trainer argument [max_steps = 100] for the full training routine and [max_steps = 10] for testing:

# config.yaml
fit:
    trainer:
        max_steps: 100
test:
    trainer:
        max_epochs: 10

now you can toggle this behavior by subcommand:

# full routine with max_steps = 100
$ python main.py --config config.yaml fit

# test only with max_epochs = 10
$ python main.py --config config.yaml test

Run from cloud yaml configs¶

For certain enterprise workloads, Lightning CLI supports running from hosted configs:

$ python main.py [subcommand] --config s3://bucket/config.yaml

For more options, refer to Remote filesystems.

Use a config via environment variables¶

For certain CI/CD systems, it’s useful to pass in raw yaml config as environment variables:

$ python main.py fit --trainer "$TRAINER_CONFIG" --model "$MODEL_CONFIG" [...]

Run from environment variables directly¶

The Lightning CLI can convert every possible CLI flag into an environment variable. To enable this, add to parser_kwargs the default_env argument:

cli = LightningCLI(..., parser_kwargs={"default_env": True})

now use the --help CLI flag with any subcommand:

$ python main.py fit --help

which will show you ALL possible environment variables that can be set:

usage: main.py [options] fit [-h] [-c CONFIG]
                            ...

optional arguments:
...
ARG:   --model.out_dim OUT_DIM
ENV:   PL_FIT__MODEL__OUT_DIM
                        (type: int, default: 10)
ARG:   --model.learning_rate LEARNING_RATE
ENV:   PL_FIT__MODEL__LEARNING_RATE
                        (type: float, default: 0.02)

now you can customize the behavior via environment variables:

# set the options via env vars
$ export PL_FIT__MODEL__LEARNING_RATE=0.01
$ export PL_FIT__MODEL__OUT_DIM=5

$ python main.py fit

Set default config files¶

To set a path to a config file of defaults, use the default_config_files argument:

cli = LightningCLI(MyModel, MyDataModule, parser_kwargs={"default_config_files": ["my_cli_defaults.yaml"]})

or if you want defaults per subcommand:

cli = LightningCLI(MyModel, MyDataModule, parser_kwargs={"fit": {"default_config_files": ["my_fit_defaults.yaml"]}})

Enable variable interpolation¶

In certain cases where multiple settings need to share a value, consider using variable interpolation. For instance:

model:
  encoder_layers: 12
  decoder_layers:
  - ${model.encoder_layers}
  - 4

To enable variable interpolation, first install omegaconf:

pip install omegaconf

Then set omegaconf when instantiating the LightningCLI class:

cli = LightningCLI(MyModel, parser_kwargs={"parser_mode": "omegaconf"})

After this, the CLI will automatically perform interpolation in yaml files:

python main.py --model.encoder_layers=12

For more details about the interpolation support and its limitations, have a look at the jsonargparse and the omegaconf documentations.

참고

There are many use cases in which variable interpolation is not the correct approach. When a parameter must always be derived from other settings, it shouldn’t be up to the CLI user to do this in a config file. For example, if the data and model both require batch_size and must be the same value, then Argument linking should be used instead of interpolation.