ttnn.experimental.conv3d

ttnn.experimental.conv3d(input_tensor: ttnn.Tensor, weight_tensor: ttnn.Tensor, config: ttnn.Conv3dConfig, *, bias_tensor: ttnn.Tensor | None, memory_config: ttnn.MemoryConfig | None, compute_kernel_config: ttnn.DeviceComputeKernelConfig | None) → ttnn.Tensor

Applies a 3D convolution over an input signal composed of several input planes. Expects Input Tensor in [N, D, H, W, C] format. Expects Weight Tensor in [1, 1, kD * kH * kW * C_in, C_out] format. Expects Bias Tensor in [1, 1, 1, 32, C_out] format. Input must be in row major interleaved format. Output will be in row major interleaved format.

Parameters:

  • input_tensor (ttnn.Tensor) – Input tensor.

  • weight_tensor (ttnn.Tensor) – Weight tensor.

  • config (ttnn.Conv3dConfig) – Configuration for the Conv3D operation.

Keyword Arguments:

  • bias_tensor (ttnn.Tensor, optional) – Bias tensor.

  • memory_config (ttnn.MemoryConfig, optional) – Memory configuration for the output of the Conv3D operation.

  • compute_kernel_config (ttnn.DeviceComputeKernelConfig, optional) – Compute kernel configuration for the Conv3D operation.

Returns:

ttnn.Tensor – Output tensor after applying the Conv3D operation.