Bunch

experimental.network_dynamics.Bunch()

Dictionary with attribute access for parameters.

A JAX PyTree-compatible parameter container that allows both dict[‘key’] and dict.key access patterns. Designed for neural dynamics parameters with support for JAX transformations.

Examples: >>> params = Bunch(a=1.0, b=2.0) >>> params.a # attribute access 1.0 >>> params[‘b’] # dict access 2.0 >>> jax.tree.map(lambda x: x * 2, params) # JAX transformations Bunch(a=2.0, b=4.0)

Methods

Name	Description
copy	Structurally copy this Bunch and any nested PyTree containers.

copy

experimental.network_dynamics.Bunch.copy()

Structurally copy this Bunch and any nested PyTree containers.

Rebuilds every container (nested Bunch, graph object, …) so mutating the copy – including plain attribute assignment like cfg.graph.delays = x – does not leak back into the original. Leaves (arrays, scalars, PRNG keys) are immutable in JAX, so sharing them is safe and avoids duplicating large arrays (graph weights, history buffers, pre-sampled noise tensors). Same operation as solve.py’s _snapshot, exposed here as .copy() so it works for ad-hoc config mutation outside prepare().

# Bunch { #tvboptim.experimental.network_dynamics.Bunch }

```python
experimental.network_dynamics.Bunch()
```

Dictionary with attribute access for parameters.

A JAX PyTree-compatible parameter container that allows both dict['key']
and dict.key access patterns. Designed for neural dynamics parameters
with support for JAX transformations.

Examples:
    >>> params = Bunch(a=1.0, b=2.0)
    >>> params.a  # attribute access
    1.0
    >>> params['b']  # dict access
    2.0
    >>> jax.tree.map(lambda x: x * 2, params)  # JAX transformations
    Bunch(a=2.0, b=4.0)

## Methods

| Name | Description |
| --- | --- |
| [copy](#tvboptim.experimental.network_dynamics.Bunch.copy) | Structurally copy this Bunch and any nested PyTree containers. |

### copy { #tvboptim.experimental.network_dynamics.Bunch.copy }

```python
experimental.network_dynamics.Bunch.copy()
```

Structurally copy this Bunch and any nested PyTree containers.

Rebuilds every container (nested Bunch, graph object, ...) so
mutating the copy -- including plain attribute assignment like
`cfg.graph.delays = x` -- does not leak back into the original.
Leaves (arrays, scalars, PRNG keys) are immutable in JAX, so
sharing them is safe and avoids duplicating large arrays (graph
weights, history buffers, pre-sampled noise tensors). Same
operation as solve.py's `_snapshot`, exposed here as `.copy()` so
it works for ad-hoc config mutation outside prepare().