1

I'm new for GPU related model training. I have Tesla C2075 with 6GB GPU and using keras CuDNNLSTM for faster training. I have installed cuda-9 with cudnn=7.0.5, tensorflow-gpu==1.12.0 and using ubuntu 16.04. For Tesla C2075 GPU model is compatible with cuda-9? I have checked https://developer.nvidia.com/cuda-gpus link in this they have mentioned tesla C2075 is compute compatible to 2.0. what is compute compatible?

And while running my model tensorflow log,

tensorflow/core/common_runtime/gpu/gpu_device.cc:1482] Ignoring visible gpu device (device: 0, name: Tesla C2075, pci bus id: 0000:03:00.0, compute capability: 2.0) with Cuda compute capability 2.0. The minimum required Cuda capability is 3.5.

And I'm also getting error while model.fit(...),

InvalidArgumentError (see above for traceback): No OpKernel was registered to support Op 'CudnnRNN' with these attrs.  Registered devices: [CPU,XLA_CPU,XLA_GPU], Registered kernels:
  device='GPU'; T in [DT_DOUBLE]
  device='GPU'; T in [DT_FLOAT]
  device='GPU'; T in [DT_HALF]

[[node bidirectional_1/CudnnRNN (defined at /usr/local/lib/python3.5/dist-packages/tensorflow/contrib/cudnn_rnn/python/ops/cudnn_rnn_ops.py:922)  = CudnnRNN[T=DT_FLOAT, direction="unidirectional", dropout=0, input_mode="linear_input", is_training=true, rnn_mode="lstm", seed=87654321, seed2=0](bidirectional_1/transpose, bidirectional_1/ExpandDims_1, bidirectional_1/ExpandDims_2, bidirectional_1/concat)]]

Thanks

Ritesh Dubal
  • 45
  • 1
  • 6

1 Answers1

3

The CUDA compute capability relates somehow to the architecture and hardware capabilities of the GPU, there is quite an extensive list in wikipedia.

The tensoflow webpage suggests that you need a GPU with CC bigger than 3.5 (older versions seemed to accept 3.0, but never lower).

Unfortunately this is a hardware limitation, the only way of changing your compute capability is using a different GPU. Simply said: you can not run Tensorflow in that GPU.

Ander Biguri
  • 35,140
  • 11
  • 74
  • 120