Ng Kam Woh

Hi authors, in the paper, you mention that nuswide "use the subset of 195,834 images that are associated with the 21 most frequent concepts", but i saw vector with size...

- added DreamCreature with arxiv, project page and github

tensorflow new version problems

1

in vgg.py, line 21, `net = tf.sub(input_tensor, mu, name="input_mean_centered")` latest version of tensorflow (1.3) should be `net = tf.subtract(input_tensor, mu, name="input_mean_centered")`

``` gold_count = len(gold_label) for idx, s in enumerate(res): if idx in gold_label and s == 1: self.res[CORRECT] += 1 self.triple_count_res[triple_count][CORRECT] += 1 if idx != self.na_id else 0 self.without_na_res[CORRECT]...

https://discuss.pytorch.org/t/saving-tensor-with-torch-save-uses-too-much-memory/46865/2 saving the text embedding as cache without .clone() will have the file size of whole tensor.

Slot attention implementation did not update the q?

1

Based on your implementation, why does the slot attention iterate with the same q & k? ``` https://github.com/wbw520/BotCL/blob/3dde3ac20cdecd7eea8c4b7cb0e04e2bb95f639b/model/contrast/slots.py#L37 def forward(self, inputs_pe, inputs, weight=None, things=None): b, n, d = inputs_pe.shape slots...

question on the cuda rasterizer

1

https://github.com/lkeab/gaussian-grouping/blob/fd359422059ef61cce19db06f63e9f3e01156a84/submodules/diff-gaussian-rasterization/cuda_rasterizer/backward.cu#L455C2-L459C59 ``` if (inside) for (int i = 0; i < C; i++) dL_dpixel[i] = dL_dpixels[i * H * W + pix_id]; for (int i = 0; i < O;...