Cannot use CRNN with Theano backend
See original GitHub issueHi,
Do you know if the pre-trained CRNN model works for any combination of library versions? I’ve tried tensorflow, but I have seen another issue here saying that it is impossible to load tensorflow weights. I’m using Keras 1.2.0 and Theano 0.9.0 and I’m getting the following error output during model.predict():
ValueError: GpuReshape: trying to reshape an array of total size 1440 into an array of total size 96. Apply node that caused the error: GpuReshape{4}(bn_0_freq_running_mean, TensorConstant{[ 1 1 96 1]}) Toposort index: 39 Inputs types: [GpuArrayType<None>(float32, (False,)), TensorType(int64, vector)] Inputs shapes: [(1440,), (4,)] Inputs strides: [(4,), (8,)] Inputs values: ['not shown', array([ 1, 1, 96, 1])] Outputs clients: [[GpuElemwise{sub,no_inplace}(GpuIncSubtensor{Set;::, ::, int64:int64:, int64:int64:}.0, GpuReshape{4}.0)]]
Issue Analytics
- State:
- Created 6 years ago
- Reactions:1
- Comments:27 (11 by maintainers)
Hi, Yes it works, in the sense that it does not crash. But, in that same sense, axis=freq and mode=1 also works (at least during prediction). However, since I want to use the pre-trained model, I should use axis=time and mode=0 since I guess that was how the code was at the time you trained the model. Is that correct?
You should do: