Gradient Modification - Githubissues

wojciechmo / deep-compression

Compress neural network with pruning and quantization using TensorFlow.

105 stars 28 forks source link

Gradient Modification #1

Open Site1997 opened 5 years ago

Site1997 commented 5 years ago

gradients_vars = optimizer.compute_gradients(loss, LAYERS_WIEGHTS)
grads = [grad for grad, var in gradients_vars]
train_step = optimizer.apply_gradients(gradients_vars)

Hi, in this code, I see that the final gradients is gradients_vars, not grads. Is this correct? (Although I see that you have a feed_dict of modified grad)

mrocr commented 5 years ago

@WojciechMormul