ttnn.conv1d

ttnn.conv1d(input_tensor: ttnn.Tensor, weight_tensor: ttnn.Tensor, device: ttnn.MeshDevice, in_channels: int, out_channels: int, batch_size: int, input_length: int, kernel_size: int, stride: int, padding: int or tuple[int, int], dilation: int, groups: int, *, bias_tensor: ttnn.Tensor | None, dtype: ttnn.DataType | None, conv_config: ttnn.Conv2dConfig | None, compute_config: ttnn.DeviceComputeKernelConfig | None, memory_config: ttnn.MemoryConfig | None, slice_config: ttnn.Conv2dSliceConfig | None, return_output_dim: bool | None, return_weights_and_bias: bool | None) → The output tensor, output length, and the preprocessed weights and bias.

Applies a 1D convolution over an input signal composed of several input planes. Implemented as a 2D Convolution of input height 1 and input width as input_length.

Parameters:

  • input_tensor (ttnn.Tensor) – The input tensor. This must be in the format [N, H, W, C]. It can be on host or device.

  • weight_tensor (ttnn.Tensor) – The weight tensor. The weights can be passed in the same format as PyTorch, [out_channels, in_channels, kernel_height, kernel_width].

  • device (ttnn.MeshDevice) – The device to use.

  • in_channels (int) – Number of input channels.

  • out_channels (int) – Number of output channels.

  • batch_size (int) – Batch size.

  • input_length (int) – Length of the input signal.

  • kernel_size (int) – Size of the convolving kernel.

  • stride (int) – Stride of the cross-correlation.

  • padding (int or tuple[int, int]) – Zero-padding added to both sides of the input. pad_length or [pad_left, pad_right].

  • dilation (int) – Spacing between kernel elements.

  • groups (int) – Number of blocked connections from input channels to output channels.

Keyword Arguments:

  • bias_tensor (ttnn.Tensor, optional) – Optional bias tensor. Default: None

  • dtype (ttnn.DataType, optional) – The data type of the input tensor. Default: None (will use the same dtype as input_tensor).

  • conv_config (ttnn.Conv2dConfig, optional) – Configuration for convolution. Default: None

  • compute_config (ttnn.DeviceComputeKernelConfig, optional) – Configuration for compute kernel. Default: None

  • memory_config (ttnn.MemoryConfig, optional) – Output Tensor’s Memory Configuration. Default: None

  • slice_config (ttnn.Conv2dSliceConfig, optional) – Configuration for slicing the input & output tensors in DRAM along the input_length (width) dimension. Use slice_type=Conv2dDRAMSliceWidth with num_slices=0 to auto-determine the number of slices, or num_slices=N to slice manually. DRAM_HEIGHT slicing is not supported because the conv1d height is always 1. If None, slicing is auto-routed by input location: inputs already in L1 run fully in L1, while DRAM/host inputs are auto width-sliced through DRAM. Default: None

  • return_output_dim (bool, optional) – If true, the op also returns the length of the output tensor. Default: False

  • return_weights_and_bias (bool, optional) – If true, the op also returns the preprocessed weight and bias on device. Default: False

Returns:

The output tensor, output length, and the preprocessed weights and bias.

  • ttnn.Tensor: The output tensor, when return_output_dim = False and return_weights_and_bias = False

  • tuple[ttnn.Tensor, int]: The output tensor, and its length, if return_output_dim = True

  • tuple[ttnn.Tensor, tuple[ttnn.Tensor, ttnn.Tensor]]: The output tensor, and its weights and biases, if return_weights_and_bias = True

  • tuple[ttnn.Tensor, int, tuple[ttnn.Tensor, ttnn.Tensor]]: The output tensor, its length, and its weights and biases, if return_output_dim = True and return_weights_and_bias = True