* hydra: enable config modules
Lets the user specify Hydra configs in installed Python modules as well
as inside directories. Adds `hydra.config_module` to `.dvc/config`.
Fixes#9740
old tests green
tests for config modules
* hydra: fix linter errors
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* hydra: fix mypy error
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* hydra: fix tests
* Update dvc/utils/hydra.py
Co-authored-by: David de la Iglesia Castro <[email protected]>
* address comments
* convince mypy not to worry
* split long line
* fix test
* check for raised error in test
---------
Co-authored-by: Dom Miketa <[email protected]>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: David de la Iglesia Castro <[email protected]>
* hydra: Use `OmegaConf.to_yaml` for dumping `.yaml` output.
Combining `OmegaConf.to_object` with our yaml DUMPER (ruaml.yaml) was not correctly handling strings like 'no'.
`OmegaConf.to_yaml` uses a custom string representer that correctly handles those cases.
Fixes#8583
* Update dvc/utils/hydra.py
Co-authored-by: Dave Berenbaum <[email protected]>
Co-authored-by: Dave Berenbaum <[email protected]>
Defaults List overrides (`defaults/foo=1,2`) were not working because of the inclusion of the `/` character.
Fixed it by updating how we construct the `merged_overrides` and added a regression test.
Requires `--queue` to be used alongside `--set-param`.
Doesn't strictly require `hydra.config.enabled`.
Add `get_hydra_sweeps` to `dvc/utils/hydra.py`. Uses hydra `BasicSweeper` to generate the different sweeps.
`glob` syntax for sweeps is not supported in this P.R.
Closes#8083
---
Using setup from https://hydra.cc/docs/tutorials/basic/your_first_app/composition/ :
```console
$ dvc exp run -S "db=mysql,postgresql" -S "schema=warehouse,school" --queue
Queueing with '{'params.yaml': ['db=mysql', 'schema=warehouse']}'.
Queued experiment '5ab98b8' for future execution.
Queueing with '{'params.yaml': ['db=mysql', 'schema=school']}'.
Queued experiment '57c2fb6' for future execution.
Queueing with '{'params.yaml': ['db=postgresql', 'schema=warehouse']}'.
Queued experiment 'b9d6391' for future execution.
Queueing with '{'params.yaml': ['db=postgresql', 'schema=school']}'.
Queued experiment '145cd55' for future execution.
```
The feature will be used depending on whether `config.hydra.enabled` is True or False.
Uses `hydra.initialize_config_dir` and `hydra.compose` (from https://hydra.cc/docs/advanced/compose_api/) to build the config and dump it to `params.yaml`. The content of the output file will be overwritten.
`config.hydra.config_dir` and `config.hydra.config_name` can be used to customize the values passed to `hydra.initialize_config_dir` and `hydra.compose`.
Can be combined with `--set-param` overrides.
Closes#8082
We don't have aiohttp support yet, due to which we have to install
without cythonized extension.
Similarly, there has been some changes in how dataclasses disallows some
mutable arguments to be set as a class property, due to which
`hydra.conf` fails to import in 3.11.
See https://github.com/python/cpython/pull/29867. For now, I have
patched the module and avoid importing as a pytest-plugin.
dvc-hdfs/pyarrow is not there yet, we'll usually see a release 2-3
months after official reelase.
dvc-hdfs is skipped in 3.11 in extras_requires. `--set-params` will
return a nice error message if hydra cannot be imported in 3.11. I
decided to only patch that in tests.
- Overriding a config value : foo.bar=value
- Appending a config value : +foo.bar=value
- Appending or overriding a config value : ++foo.bar=value
- Removing a config value : ~foo.bar, ~foo.bar=value
See https://hydra.cc/docs/advanced/override_grammar/basic/#modifying-the-config-object
---
Breaking changes:
To modify a list, `foo[0]=bar` must now be passed as `foo.0=bar`.
Modifying a nested list inside a dictionary is not supported by omegaconf.
---
Closes#4883Closes#5868Closes#6129
The tests should be in dvc-objects repo npw, and will be migrated soon.
The nanotime and diskcache are specified in dvc-objects, so we don't need
those here.
By splitting dvc.cli into a module, we'll be able to share cli-related
utilities, split parser-related stuffs into smaller parts, improve
reusability and improve API for command extensions in the future.
skshetry
Saugat Pachhai (सौगात)
Daniele Trifirò
Paweł Redzyński
Ruslan Kuprieiev