ValueError: Negative dimension size caused by subtracting 3 from 2 for 'MaxPool' (op: 'MaxPool') with input shapes: [?,2,224,64].

iamhankai / attribute-aware-attention

[ACM MM 2018] Attribute-Aware Attention Model for Fine-grained Representation Learning

https://arxiv.org/abs/1901.00392

156 stars 30 forks source link

ValueError: Negative dimension size caused by subtracting 3 from 2 for 'MaxPool' (op: 'MaxPool') with input shapes: [?,2,224,64]. #13

Open zlxhhh opened 5 years ago

zlxhhh commented 5 years ago

when i run run.sh.it happened.what should i do?

iamhankai commented 5 years ago

refer to https://github.com/iamhankai/attribute-aware-attention/issues/12#issuecomment-494344421

zlxhhh commented 5 years ago

thank you for your answer

zlxhhh commented 5 years ago

how to change image_data_format change channels_last to channels_first

iamhankai commented 5 years ago

Yes

EpilogueCc commented 5 years ago

I have changed image_data_format in keras.json but NO useful，ValueError: Negative dimension size caused by subtracting 2 from 1 for 'MaxPool_1' (op: 'MaxPool') with input shapes: [?,1,224,128].

iamhankai commented 5 years ago

Try different image_data_format and backend.

gengyanlei commented 5 years ago

Hello, author. What is the version of the tensorflow framework that your keras depends on?

EpilogueCc commented 5 years ago

tryed，but no useful. Can you tell me which the specific version of keras and tensorflow？

iamhankai commented 5 years ago

@gengyanlei @EpilogueCc

I used

Keras 1.2.1 ("image_data_format": "channels_first")
Theano backend (I forget the version)

EpilogueCc commented 5 years ago

which python version u use , I cannot run by gpu ERROR (theano.gpuarray): pygpu was configured but could not be imported or is too old (version 0.7 or higher required)

iamhankai commented 5 years ago

Python2.7