nshconfig

https://nima.sh/nshconfig/
Admin
nshconfig is a Python configuration library providing typed drafts, live interpolation, and...
Tokens:5,204
Snippets:78
Trust Score:6.1
Update:3 weeks ago
Show doc for...
Tokens:
Raw
### Install nshconfig v3 alpha

Source: https://nima.sh/nshconfig/_sources/installation.md.txt

Install the nshconfig v3 alpha release. Supports Python 3.10-3.14 and Pydantic 2.13 through the latest 2.x. For trusted notebook transport, install nshconfig[transport] with the same version constraint.

```bash
pip install 'nshconfig==3.0.0a0'
```

--------------------------------

### Example project directory structure

Source: https://nima.sh/nshconfig/_sources/guides/project-layout.md.txt

Shows a recommended layout with schemas.py, presets.py, experiments/wide.py, and train.py. Use ordinary Python modules for schemas, preset helpers, experiments, and execution.

```text
project/
    schemas.py
    presets.py
    experiments/
        wide.py
    train.py
```

--------------------------------

### Serialize and restore config with cloudpickle

Source: https://nima.sh/nshconfig/_sources/guides/transport.md.txt

Use this snippet to serialize a draft configuration and its finalized version with cloudpickle, then restore both objects. Requires the optional `transport` extra to be installed. The restored objects preserve the editable/final distinction, interpolation rules, and child ownership.

```python
import cloudpickle

payload = cloudpickle.dumps((draft, draft.finalize()))
restored_draft, restored_final = cloudpickle.loads(payload)
```

--------------------------------

### Define config with interpolation and draft/finalize workflow

Source: https://nima.sh/nshconfig/_sources/quickstart.md.txt

Shows how to define a config with an interpolated field that references the root config's width, create an editable draft, finalize it, and verify that the interpolated value is frozen after finalization. Note that .draft() allows required fields to be unset, while normal constructors statically check required arguments.

```python
import nshconfig as C

class Model(C.Config):
    width: int = C.interp(lambda c: c.root(Run).width)

class Run(C.Config):
    name: str
    width: int = 128
    model: Model = Model()

config = Run.draft()
config.name = "example"
config.width = 256
assert config.model.width == 256
ready = config.finalize()
config.width = 512
assert ready.model.width == 256
```

--------------------------------

### Define config classes and use draft/finalize with interpolation

Source: https://nima.sh/nshconfig/quickstart.html

Defines Model and Run config classes. Model.width interpolates from the root Run config. Creates a draft, sets required fields, finalizes, and verifies the interpolated value. Note that .draft() allows unset required fields, but reading them fails at runtime.

```python
import nshconfig as C

class Model(C.Config):
    width: int = C.interp(lambda c: c.root(Run).width)

class Run(C.Config):
    name: str
    width: int = 128
    model: Model = Model()

config = Run.draft()
config.name = "example"
config.width = 256
assert config.model.width == 256
ready = config.finalize()
config.width = 512
assert ready.model.width == 256

```

--------------------------------

### Dispatch publish workflow with gh

Source: https://nima.sh/nshconfig/guides/publishing.html

Manually dispatch the publish workflow on the main branch with the exact version to release. The version must match pyproject.toml and the workflow will fail if CI has not passed or the version is invalid.

```bash
gh workflow run publish.yml --repo nimashoghi/nshconfig --ref main -f version=3.0.0a0
```
Tokens:
Raw