Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

dimension of sample_inputs and self.inputs

See original GitHub issue

I got a problem that

    tf.app.run()
  File "/home/jd730/p3/lib/python3.5/site-packages/tensorflow/python/platform/app.py", line 48, in run
    _sys.exit(main(_sys.argv[:1] + flags_passthrough))
  File "main.py", line 114, in main
    dcgan.train(FLAGS)
  File "/home/jd730/Creative-Adversarial-Networks/model.py", line 402, in train
    self.y:sample_labels,
  File "/home/jd730/p3/lib/python3.5/site-packages/tensorflow/python/client/session.py", line 895, in run
    run_metadata_ptr)
  File "/home/jd730/p3/lib/python3.5/site-packages/tensorflow/python/client/session.py", line 1100, in _run
    % (np_val.shape, subfeed_t.name, str(subfeed_t.get_shape())))
ValueError: Cannot feed value of shape (64, 64, 64, 3) for Tensor 'real_images:0', which has shape '(36, 64, 64, 3)'

I think its because difference of dimension of self.inputs and samples_inputs. In model.py

According to line 117, self.inputs should be [batch_size, image_dims(ex. 64 64 3)] self.inputs = tf.placeholder(tf.float32, [self.batch_size] + image_dims, name='real_images')

However, on line 237, sample_inputs which is a factor of self.inputs on line 398 is defined by

      sample = [
          get_image(sample_file,
                    input_height=self.input_height,
                    input_width=self.input_width,
                    resize_height=self.output_height,
                    resize_width=self.output_width,
                    crop=self.crop,
                    grayscale=self.grayscale) for sample_file in sample_files]
      if (self.grayscale):
        sample_inputs = np.array(sample).astype(np.float32)[:, :, :, None]
      else:
        sample_inputs = np.array(sample).astype(np.float32)

So, its dimension is [sample_num(=sample_size), image_dim] .

I thought it should change to batch_size rather sample_num. But, carpedm20’s DCGAN also used that code. May I ask if you can explain this situation?

Issue Analytics

State:
Created 6 years ago
Comments:8 (2 by maintainers)

Top GitHub Comments

1reaction

phillip-kravtsovcommented, Nov 8, 2017

If you pull the new code from the repository, which has been updated since you asked your question, and run train.sh with different values for batch_size and sample_size, the code should not throw an error and should behave as expected. There shouldn’t be any dimensionality issues (there weren’t when I tested it).

The original DCGAN code assumed that batch_size and sample_size were the same.

0reactions

jd730commented, Apr 12, 2019

@galoisgroupcn Hi, I am good.

Thank you. Jaedong

Top Results From Across the Web

deepmask/TrainerSharpMask.lua at master - GitHub

function: copy inputs/labels to CUDA tensor. function Trainer:copySamples(sample). self.inputs:resize(sample.inputs:size()):copy(sample.inputs).

How to run a trained PyTorch Model on my sample inputs?

The problem with shape is that the model expects input of shape [BatchSize, Channel, Height, Width] and the current input was [Channel, ...

Source code for GPy.testing.model_tests

[docs] def test_setXY(self): m = GPy.models.GPRegression(self.X, self.Y) m. ... 1 dimensional example # sample inputs and outputs self.

A joint learning framework for Gaussian processes regression ...

In this paper, we propose a novel GPR model, where multi-dimensional sample inputs are viewed as signals generated over an underlying graph.

| notebook.community

image is normalized CIELAB, depth is not normalized. self.image, self.depth ... entropy): labels.fill(0) for b in range(batch_size): self.sample(inputs, ...