zalandoresearch / psgan

Periodic Spatial Generative Adversarial Networks
MIT License
140 stars 33 forks source link

Image generating problem #8

Open PJeys opened 2 years ago

PJeys commented 2 years ago

Hello everybody! I am trying to generate some images from pretrained models to see how it works in general. But after calling python2 demo_generation.py models/honey_filters64_npx161_5gL_5dL_60Global_3Periodic_FalseAffine_30Local_epoch100.psgan

I'm facing an error:

using stored model models/honey_filters64_npx161_5gL_5dL_60Global_3Periodic_FalseAffine_30Local_epoch100.psgan loading parameters from file: models/honey_filters64_npx161_5gL_5dL_60Global_3Periodic_FalseAffine_30Local_epoch100.psgan global dimensions of loaded config file 60 Compiling the network... Discriminator done. (took 13.628841877 s) Generator done. (took 7.85284900665 s) generate function done. (took 2.22278308868 s) nz 96 global Dimensions 60 periodic Dimensions 3 G values [512, 256, 128, 64, 3] [(5, 5), (5, 5), (5, 5), (5, 5), (5, 5)] D values [3, 64, 128, 256, 512] [(5, 5), (5, 5), (5, 5), (5, 5), (5, 5)] Traceback (most recent call last): File "demo_generation.py", line 70, in sample_texture(psgan) File "demo_generation.py", line 61, in sample_texture data = psgan.generate(z_sample) File "/Users/michaelmedved/Library/Python/2.7/lib/python/site-packages/theano/compile/function_module.py", line 886, in call storage_map=getattr(self.fn, 'storage_map', None)) File "/Users/michaelmedved/Library/Python/2.7/lib/python/site-packages/theano/gof/link.py", line 325, in raise_with_op reraise(exc_type, exc_value, exc_trace) File "/Users/michaelmedved/Library/Python/2.7/lib/python/site-packages/theano/compile/function_module.py", line 873, in call self.fn() if output_subset is None else\ AssertionError: AbstractConv_gradInputs shape mismatch: shape of filters does not match given kshp. Apply node that caused the error: Assert{msg='AbstractConv_gradInputs shape mismatch: shape of filters does not match given kshp.'}(W, Elemwise{eq,no_inplace}.0, Elemwise{eq,no_inplace}.0, Elemwise{eq,no_inplace}.0, Elemwise{eq,no_inplace}.0) Toposort index: 72 Inputs types: [TensorType(float64, 4D), TensorType(bool, scalar), TensorType(bool, scalar), TensorType(bool, scalar), TensorType(bool, scalar)] Inputs shapes: [(96, 512, 5, 5), (), (), (), ()] Inputs strides: [(102400, 200, 40, 8), (), (), (), ()] Inputs values: ['not shown', array(False), array(True), array(True), array(True)] Inputs type_num: [12, 0, 0, 0, 0] Outputs clients: [[Subtensor{::, ::, ::int64, ::int64}(Assert{msg='AbstractConv_gradInputs shape mismatch: shape of filters does not match given kshp.'}.0, Constant{-1}, Constant{-1})]]

Backtrace when the node is created(use Theano flag traceback.limit=N to make it longer): File "demo_generation.py", line 64, in psgan = PSGAN(name=name) File "/Users/michaelmedved/Development/kaleidoscope/psgan/psgan/psgan.py", line 174, in init self._build_sgan() File "/Users/michaelmedved/Development/kaleidoscope/psgan/psgan/psgan.py", line 327, in _build_sgan prediction_gen = lasagne.layers.get_output(gen_X) File "/Users/michaelmedved/Library/Python/2.7/lib/python/site-packages/lasagne/layers/helper.py", line 197, in get_output all_outputs[layer] = layer.get_output_for(layer_inputs, kwargs) File "/Users/michaelmedved/Library/Python/2.7/lib/python/site-packages/lasagne/layers/conv.py", line 352, in get_output_for conved = self.convolve(input, kwargs) File "/Users/michaelmedved/Library/Python/2.7/lib/python/site-packages/lasagne/layers/conv.py", line 981, in convolve conved = op(self.W, input, output_size)

Debugprint of the apply node: Assert{msg='AbstractConv_gradInputs shape mismatch: shape of filters does not match given kshp.'} [id A] <TensorType(float64, 4D)> ''
|W [id B] <TensorType(float64, 4D)> |Elemwise{eq,no_inplace} [id C] <TensorType(bool, scalar)> ''
| |Shape_i{0} [id D] <TensorType(int64, scalar)> ''
| | |W [id B] <TensorType(float64, 4D)> | |TensorConstant{102} [id E] <TensorType(int8, scalar)> |Elemwise{eq,no_inplace} [id F] <TensorType(bool, scalar)> ''
| |Shape_i{1} [id G] <TensorType(int64, scalar)> ''
| | |W [id B] <TensorType(float64, 4D)> | |TensorConstant{512} [id H] <TensorType(int16, scalar)> |Elemwise{eq,no_inplace} [id I] <TensorType(bool, scalar)> ''
| |Shape_i{2} [id J] <TensorType(int64, scalar)> ''
| | |W [id B] <TensorType(float64, 4D)> | |TensorConstant{5} [id K] <TensorType(int8, scalar)> |Elemwise{eq,no_inplace} [id L] <TensorType(bool, scalar)> ''
|Shape_i{3} [id M] <TensorType(int64, scalar)> ''
| |W [id B] <TensorType(float64, 4D)> |TensorConstant{5} [id K] <TensorType(int8, scalar)>

Storage map footprint:

Can you explain please where is the problem and how can it be fixed