neuraltrain.models.conv_transformer.ConvTransformerModel

class neuraltrain.models.conv_transformer.ConvTransformerModel(in_channels: int, out_channels: int | None, config: ConvTransformer)

nn.Module implementation of ConvTransformer.

forward(x: Tensor, subject_ids: Tensor | None = None, channel_positions: Tensor | None = None, neuro_device_type: str | None = None) → dict[str, Tensor]

Forward pass through encoder, optional transformer, and output layer.

Parameters:

• x (Tensor) – Input of shape (B, C, T).

• subject_ids (Tensor or None) – Per-example subject indices, shape (B,).

• channel_positions (Tensor or None) – Normalised electrode coordinates, shape (B, C, D).

• neuro_device_type (str or None) – Name of device represented in the batch (e.g. "Eeg", "Meg"). If None, the device embedding is not applied.