2

I am using anaconda python 3.7 and tensorflow 2.1 with cuda 10.1 and cudnn 7.6.5, and trying to run the retinaset (https://github.com/fizyr/keras-retinanet): 

python keras_retinanet/bin/train.py --freeze-backbone --random-transform --batch-size 8 --steps 500 --epochs 10 csv annotations.csv classes.csv

Here below are the resultant errors:

Epoch 1/10
2020-02-10 20:34:37.807590: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudnn64_7.dll
2020-02-10 20:34:38.835777: E tensorflow/stream_executor/cuda/cuda_dnn.cc:329] Could not create cudnn handle: CUDNN_STATUS_INTERNAL_ERROR
2020-02-10 20:34:39.753051: E tensorflow/stream_executor/cuda/cuda_dnn.cc:329] Could not create cudnn handle: CUDNN_STATUS_INTERNAL_ERROR
2020-02-10 20:34:39.776706: W tensorflow/core/common_runtime/base_collective_executor.cc:217] BaseCollectiveExecutor::StartAbort Unknown: Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so try looking to see if a warning log message was printed above.
         [[{{node conv1/convolution}}]]
Traceback (most recent call last):
  File "keras_retinanet/bin/train.py", line 530, in <module>
    main()
  File "keras_retinanet/bin/train.py", line 525, in main
    initial_epoch=args.initial_epoch
  File "C:\Anaconda\Anaconda3.7\lib\site-packages\keras\legacy\interfaces.py", line 91, in wrapper
    return func(*args, **kwargs)
  File "C:\Anaconda\Anaconda3.7\lib\site-packages\keras\engine\training.py", line 1732, in fit_generator
    initial_epoch=initial_epoch)
  File "C:\Anaconda\Anaconda3.7\lib\site-packages\keras\engine\training_generator.py", line 220, in fit_generator
    reset_metrics=False)
  File "C:\Anaconda\Anaconda3.7\lib\site-packages\keras\engine\training.py", line 1514, in train_on_batch
    outputs = self.train_function(ins)
  File "C:\Anaconda\Anaconda3.7\lib\site-packages\tensorflow_core\python\keras\backend.py", line 3727, in __call__
    outputs = self._graph_fn(*converted_inputs)
  File "C:\Anaconda\Anaconda3.7\lib\site-packages\tensorflow_core\python\eager\function.py", line 1551, in __call__
    return self._call_impl(args, kwargs)
  File "C:\Anaconda\Anaconda3.7\lib\site-packages\tensorflow_core\python\eager\function.py", line 1591, in _call_impl
    return self._call_flat(args, self.captured_inputs, cancellation_manager)
  File "C:\Anaconda\Anaconda3.7\lib\site-packages\tensorflow_core\python\eager\function.py", line 1692, in _call_flat
    ctx, args, cancellation_manager=cancellation_manager))
  File "C:\Anaconda\Anaconda3.7\lib\site-packages\tensorflow_core\python\eager\function.py", line 545, in call
    ctx=ctx)
  File "C:\Anaconda\Anaconda3.7\lib\site-packages\tensorflow_core\python\eager\execute.py", line 67, in quick_execute
    six.raise_from(core._status_to_exception(e.code, message), None)
  File "<string>", line 3, in raise_from
tensorflow.python.framework.errors_impl.UnknownError:  Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so try looking to see if a warning log message was printed above.
         [[node conv1/convolution (defined at C:\Anaconda\Anaconda3.7\lib\site-packages\keras\backend\tensorflow_backend.py:3009) ]] [Op:__inference_keras_scratch_graph_12376]

Function call stack:
keras_scratch_graph

Anyone has experienced similar problems?

jwm
  • 4,832
  • 10
  • 46
  • 78

4 Answers4

7

I was getting the same error when trying to train my CNN model on two GPUs using tf.distribute.MirroredStrategy(). I found a workaround for now that allows me to use both of them (though training on a single GPU worked just fine). Try putting the following at the beginning of your application:

config = tf.compat.v1.ConfigProto()
config.gpu_options.allow_growth = True
session =tf.compat.v1.InteractiveSession(config=config)

Hope that helps!

Joe Mayo
  • 7,501
  • 7
  • 41
  • 60
Gary
  • 71
  • 4
  • I added these lines to my script, and the same "convolution" error still occurs. Do I have to put it on top of the main script, or the training script, or both btw? – Leevo May 08 '20 at 15:06
1

Do this:

physical_devices = tf.config.experimental.list_physical_devices(‘GPU’)
tf.config.experimental.set_memory_growth(physical_devices[0], True)
סטנלי גרונן
  • 2,917
  • 23
  • 46
  • 68
Atif Faridi
  • 11
  • 1
0

According to this comment in a Tensorflow GitHub issue, this error can be caused by your GPU's memory limit being hit (you can check GPU usage using the commands nvidia-smi or gpustat).

If setting tf.config.experimental.set_memory_growth = True does not work, hopefully limiting GPU memory usage manually works:

gpus = tf.config.experimental.list_physical_devices('GPU')
if gpus:
    # Restrict TensorFlow to only allocate 1GB * 2 of memory on the first GPU
    try:
        tf.config.experimental.set_virtual_device_configuration(
            gpus[0],
            [tf.config.experimental.VirtualDeviceConfiguration(memory_limit=1024 * 2)])
        logical_gpus = tf.config.experimental.list_logical_devices('GPU')
        print(len(gpus), "Physical GPUs,", len(logical_gpus), "Logical GPUs")
    except RuntimeError as e:
        # Virtual devices must be set before GPUs have been initialized
        print(e)

Credit goes to BryanBo-Cao for his comment.

NJHJ
  • 142
  • 3
  • 9
0

Got the same error with python 3.7.9, tensorflow 2.1.0, cuda 10.1.105, and cudnn 7.6.5. Solved after updating GPU driver from NVIDIA.

xldk
  • 11
  • 2