search

ychfan / tf_estimator_barebone

TensorFlow template with high-level API
MIT License
70 stars 54 forks source link

Training with my own dataset #4

Closed bensuucar closed 5 years ago

bensuucar commented 5 years ago

Hi,

I would like to train EDSR/WDSR with my own training data from scratch. However I am facing with this issue which I couldnt find any solution for.

tensorflow.python.framework.errors_impl.InvalidArgumentError: Expected size[0] in [0, 108], but got 144 [[Node: Slice_1 = Slice[Index=DT_INT32, T=DT_UINT8, _class=["loc:@cond/Switch_2"], _device="/device:CPU:0"](arg1, Slice_1/begin, Slice_1/size)]] [[Node: IteratorGetNext = IteratorGetNextoutput_shapes=[[16,48,48,3], [16,144,144,3]], output_types=[DT_FLOAT, DT_FLOAT], _device="/job:localhost/replica:0/task:0/device:CPU:0"]]

capture

Could you please help?

ychfan commented 5 years ago

Could you share more details about your code? I guess the problem is from func list_files. It may caused by the lr and hr image filenames are not aligned. But I am not sure how you implement the input_fn