def upsample_conv_2d()

in jcm/models/up_or_down_sampling.py [0:0]

• 31 lines of code
• 4 McCabe index (conditional complexity)

def upsample_conv_2d(x, w, k=None, factor=2, gain=1, data_format="NHWC"):
    """Fused `upsample_2d()` followed by `tf.nn.conv2d()`.

    Padding is performed only once at the beginning, not between the
    operations.
    The fused op is considerably more efficient than performing the same
    calculation
    using standard TensorFlow ops. It supports gradients of arbitrary order.
    Args:
      x:            Input tensor of the shape `[N, C, H, W]` or `[N, H, W,
        C]`.
      w:            Weight tensor of the shape `[filterH, filterW, inChannels,
        outChannels]`. Grouped convolution can be performed by `inChannels =
        x.shape[0] // numGroups`.
      k:            FIR filter of the shape `[firH, firW]` or `[firN]`
        (separable). The default is `[1] * factor`, which corresponds to
        nearest-neighbor upsampling.
      factor:       Integer upsampling factor (default: 2).
      gain:         Scaling factor for signal magnitude (default: 1.0).
      data_format:  `'NCHW'` or `'NHWC'` (default: `'NCHW'`).

    Returns:
      Tensor of the shape `[N, C, H * factor, W * factor]` or
      `[N, H * factor, W * factor, C]`, and same datatype as `x`.
    """

    assert isinstance(factor, int) and factor >= 1

    # Check weight shape.
    assert len(w.shape) == 4
    convH = w.shape[0]
    convW = w.shape[1]
    inC = w.shape[2]
    outC = w.shape[3]
    assert convW == convH

    # Setup filter kernel.
    if k is None:
        k = [1] * factor
    k = _setup_kernel(k) * (gain * (factor**2))
    p = (k.shape[0] - factor) - (convW - 1)

    stride = [factor, factor]
    # Determine data dimensions.
    if data_format == "NCHW":
        num_groups = _shape(x, 1) // inC
    else:
        num_groups = _shape(x, 3) // inC

    # Transpose weights.
    w = jnp.reshape(w, [convH, convW, inC, num_groups, -1])
    w = jnp.transpose(w[::-1, ::-1], [0, 1, 4, 3, 2])
    w = jnp.reshape(w, [convH, convW, -1, num_groups * inC])

    ## Original TF code.
    # x = tf.nn.conv2d_transpose(
    #     x,
    #     w,
    #     output_shape=output_shape,
    #     strides=stride,
    #     padding='VALID',
    #     data_format=data_format)
    ## JAX equivalent
    x = jax.lax.conv_transpose(
        x,
        w,
        strides=stride,
        padding="VALID",
        transpose_kernel=True,
        dimension_numbers=(data_format, "HWIO", data_format),
    )

    return _simple_upfirdn_2d(
        x, k, pad0=(p + 1) // 2 + factor - 1, pad1=p // 2 + 1, data_format=data_format
    )