Default upsampling behavior when mode=bilinear is changed to align_corners=False since 0.4.0. Please specify align_corners=True if the old behavior is desired about mlab HOT 1 OPEN

// reference: interpolate in torch/nn/functional.py
// size can be none, int or intlist
// scale_factors can be none, float, or floatlist
at::Tensor interpolate(
    const at::Tensor& input,
    const IValue& size,
    const IValue& scale_factors,
    const std::string& mode,
    c10::optional<bool> align_corners,
    c10::optional<bool> recompute_scale_factor) {
  if ((mode == "nearest" || mode == "area")) {
    if (align_corners != c10::nullopt) {
      throw std::runtime_error(
          "align_corners option can only be set with the "
          "interpolating modes: linear | bilinear | bicubic | trilinear");
    }
  } else {
    if (align_corners == c10::nullopt) {
      TORCH_WARN(
          "Default upsampling behavior when mode=",
          mode,
          " is changed "
          "to align_corners=False since 0.4.0. Please specify align_corners=True "
          "if the old behavior is desired. See the documentation of nn.Upsample for details");
      align_corners = false;
    }
  }