# `Tinkex.Future.Combiner`
[🔗](https://github.com/North-Shore-AI/tinkex/blob/v0.4.0/lib/tinkex/future/combiner.ex#L1)

Helpers for combining chunked forward/backward results returned by the API.

Training client callers rely on this module to flatten chunk responses and
apply metric reduction identical to the Python SDK.

# `combine_forward_backward_results`

```elixir
@spec combine_forward_backward_results([Tinkex.Types.ForwardBackwardOutput.t()]) ::
  Tinkex.Types.ForwardBackwardOutput.t()
```

Combine multiple chunked `%ForwardBackwardOutput{}` structs into a single
struct.

* `loss_fn_output_type` is taken from the first chunk. When later chunks
  disagree a warning is logged but the first value still wins.
* `loss_fn_outputs` are flattened in chunk order.
* `metrics` are merged via `Tinkex.MetricsReduction.reduce/1`.

---

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