Closed crosofg closed 8 years ago
Seems like you used Theano. I test on cpu/mac, cpu/ubuntu, gpu/ubuntu, cuDNN 5005. Could you update your CUDA/theano/keras?
On 9Sep 2016, at 09:09, crosofg notifications@github.com wrote:
File "example.py", line 62, in main model = AudioConvnet() File "/home/ubuntu/gg_code/music-auto_tagging-keras/audio_convnet.py", line 77, in AudioConvnet x = BatchNormalization(axis=time_axis, name='bn_0_freq')(melgram_input) File "/home/ubuntu/anaconda2/lib/python2.7/site-packages/keras/engine/topology.py", line 515, in call self.add_inbound_node(inbound_layers, node_indices, tensor_indices) File "/home/ubuntu/anaconda2/lib/python2.7/site-packages/keras/engine/topology.py", line 573, in add_inbound_node Node.create_node(self, inbound_layers, node_indices, tensor_indices) File "/home/ubuntu/anaconda2/lib/python2.7/site-packages/keras/engine/topology.py", line 150, in create_node output_tensors = to_list(outbound_layer.call(input_tensors[0], mask=input_masks[0])) File "/home/ubuntu/anaconda2/lib/python2.7/site-packages/keras/layers/normalization.py", line 131, in call epsilon=self.epsilon) File "/home/ubuntu/anaconda2/lib/python2.7/site-packages/keras/backend/theano_backend.py", line 392, in batch_normalization 'spatial', epsilon) File "/home/ubuntu/anaconda2/lib/python2.7/site-packages/theano/sandbox/cuda/dnn.py", line 2770, in dnn_batch_normalization_test (gamma.ndim, beta.ndim, ndim)) ValueError: gamma and beta must be of the same dimensionality as inputs; got 1 and 1 instead of 4 Is there something wrong with the setup?
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/keunwoochoi/music-auto_tagging-keras/issues/1, or mute the thread https://github.com/notifications/unsubscribe-auth/APZ8xdtmPXzt-aj6cg0n7Tafso9AuLOoks5qoVqRgaJpZM4J5B-V.
#define CUDNN_MAJOR 5
#define CUDNN_MINOR 1
#define CUDNN_PATCHLEVEL 3
--
#define CUDNN_VERSION (CUDNN_MAJOR * 1000 + CUDNN_MINOR * 100 + CUDNN_PATCHLE VEL)
Using gpu device 0: GRID K520 (CNMeM is disabled, cuDNN 5103)
>>> print theano.__version__
0.9.0dev2.dev-4b97b62000aa025c79e7f492bdf1a1de6fdb5b8f
>>> print keras.__version__
1.0.8
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2015 NVIDIA Corporation
Built on Tue_Aug_11_14:27:32_CDT_2015
Cuda compilation tools, release 7.5, V7.5.17
These are the versions I am using.
I don’t think I can understand the exact problem. cuDNN version is slightly different and I am using CUDA 8.0, not 7.5.., apologise that this is all I can guess. But it must be something about cuda according to the error message.
On 9Sep 2016, at 11:31, crosofg notifications@github.com wrote:
define CUDNN_MAJOR 5
define CUDNN_MINOR 1
define CUDNN_PATCHLEVEL 3
define CUDNN_VERSION (CUDNN_MAJOR * 1000 + CUDNN_MINOR * 100 + CUDNN_PATCHLE VEL)
Using gpu device 0: GRID K520 (CNMeM is disabled, cuDNN 5103)
print theano.version 0.9.0dev2.dev-4b97b62000aa025c79e7f492bdf1a1de6fdb5b8f print keras.version 1.0.8
nvcc: NVIDIA (R) Cuda compiler driver Copyright (c) 2005-2015 NVIDIA Corporation Built on Tue_Aug_11_14:27:32_CDT_2015 Cuda compilation tools, release 7.5, V7.5.17 These are the versions I am using.
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/keunwoochoi/music-auto_tagging-keras/issues/1#issuecomment-245948156, or mute the thread https://github.com/notifications/unsubscribe-auth/APZ8xaBYzqykh4aW9s9mJvh9ABAjU8BWks5qoXu6gaJpZM4J5B-V.
Probably could you learn it on CPU just for a test?
On 9Sep 2016, at 11:31, crosofg notifications@github.com wrote:
define CUDNN_MAJOR 5
define CUDNN_MINOR 1
define CUDNN_PATCHLEVEL 3
define CUDNN_VERSION (CUDNN_MAJOR * 1000 + CUDNN_MINOR * 100 + CUDNN_PATCHLE VEL)
Using gpu device 0: GRID K520 (CNMeM is disabled, cuDNN 5103)
print theano.version 0.9.0dev2.dev-4b97b62000aa025c79e7f492bdf1a1de6fdb5b8f print keras.version 1.0.8
nvcc: NVIDIA (R) Cuda compiler driver Copyright (c) 2005-2015 NVIDIA Corporation Built on Tue_Aug_11_14:27:32_CDT_2015 Cuda compilation tools, release 7.5, V7.5.17 These are the versions I am using.
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/keunwoochoi/music-auto_tagging-keras/issues/1#issuecomment-245948156, or mute the thread https://github.com/notifications/unsubscribe-auth/APZ8xaBYzqykh4aW9s9mJvh9ABAjU8BWks5qoXu6gaJpZM4J5B-V.
its working with CPU. I wonder what is off with GPU config? Everything else is fine.
Yes must be something about GPU configuration. I can't find any better answer than that..
@crosofg, in case you haven't figured this out yet, it's a Keras issue. Pull the bleeding edge from the Keras git repo to resolve.
Is there something wrong with the setup?