Downgrade CUDA in DSVM Ubuntu RRS feed

  • Question

  • I am having a DSVM with UBUNTU image. Is it possible to downgrade the CUDA version from 10.1 to 10.0, I have manually installed the same but the CUDNN test is showing matching the following error:

    cudnnGetVersion() : 7605 , CUDNN_VERSION from cudnn.h : 7605 (7.6.5)
    Cuda failurer version : GCC 7.5.0
    Error: no CUDA-capable device is detected

    and the tensorflow device placement is giving:

    2020-06-23 22:34:54.560257: I tensorflow/core/platform/cpu_feature_guard.cc:142] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA
    2020-06-23 22:34:54.589600: I tensorflow/core/platform/profile_utils/cpu_utils.cc:94] CPU Frequency: 2593985000 Hz
    2020-06-23 22:34:54.591484: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x5643c1b71610 initialized for platform Host (this does not guarantee that XLA will be used). Devices:
    2020-06-23 22:34:54.591518: I tensorflow/compiler/xla/service/service.cc:176]   StreamExecutor device (0): Host, Default Version
    2020-06-23 22:34:54.597040: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcuda.so.1
    2020-06-23 22:34:54.617355: E tensorflow/stream_executor/cuda/cuda_driver.cc:318] failed call to cuInit: CUDA_ERROR_NO_DEVICE: no CUDA-capable device is detected
    2020-06-23 22:34:54.617423: I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:156] kernel driver does not appear to be running on this host (MachinelearningProdVM): /proc/driver/nvidia/version does not exist
    Device mapping:
    /job:localhost/replica:0/task:0/device:XLA_CPU:0 -> device: XLA_CPU device
    2020-06-23 22:34:54.618237: I tensorflow/core/common_runtime/direct_session.cc:359] Device mapping:
    /job:localhost/replica:0/task:0/device:XLA_CPU:0 -> device: XLA_CPU device
    Please help rectify this issue, I need to use Tensorflow 1.15.0

    Tuesday, June 23, 2020 10:48 PM