# `Tinkex.Types.TypeAliases`
[🔗](https://github.com/North-Shore-AI/tinkex/blob/v0.4.0/lib/tinkex/types/type_aliases.ex#L1)

Type aliases for Python SDK parity.

These type aliases mirror Python's TypeAlias definitions for
union types and dictionary types.

## ModelInputChunk

Union of input chunk types that can appear in model input:
- `EncodedTextChunk` - tokenized text
- `ImageAssetPointerChunk` - reference to image asset
- `ImageChunk` - inline image data

Mirrors Python `tinker.types.ModelInputChunk`.

## LossFnInputs / LossFnOutput

Dictionary mapping string keys to TensorData values.
Used for loss function inputs and outputs.

Mirrors Python `tinker.types.LossFnInputs` and `tinker.types.LossFnOutput`.

# `loss_fn_inputs`

```elixir
@type loss_fn_inputs() :: %{optional(String.t()) => Tinkex.Types.TensorData.t()}
```

Dictionary mapping string keys to TensorData.

Mirrors Python `LossFnInputs = Dict[str, TensorData]`.

# `loss_fn_output`

```elixir
@type loss_fn_output() :: %{optional(String.t()) => Tinkex.Types.TensorData.t()}
```

Dictionary mapping string keys to TensorData.

Mirrors Python `LossFnOutput = Dict[str, TensorData]`.

# `model_input_chunk`

```elixir
@type model_input_chunk() ::
  Tinkex.Types.EncodedTextChunk.t()
  | Tinkex.Types.ImageAssetPointerChunk.t()
  | Tinkex.Types.ImageChunk.t()
```

Union of model input chunk types.

Mirrors Python `ModelInputChunk = Union[EncodedTextChunk, ImageAssetPointerChunk, ImageChunk]`.

---

*Consult [api-reference.md](api-reference.md) for complete listing*