Tensorflow 因 CUBLAS_STATUS_ALLOC_FAILED 崩溃

Question

我使用简单的 MINST 神经网络程序在 Windows 10 上运行 tensorflow-gpu。当它尝试运行时，会遇到 CUBLAS_STATUS_ALLOC_FAILED 错误。谷歌搜索没有找到任何结果。

I c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\core\common_runtime\gpu\gpu_device.cc:885] Found device 0 with properties:
name: GeForce GTX 970
major: 5 minor: 2 memoryClockRate (GHz) 1.253
pciBusID 0000:0f:00.0
Total memory: 4.00GiB
Free memory: 3.31GiB
I c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\core\common_runtime\gpu\gpu_device.cc:906] DMA: 0
I c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\core\common_runtime\gpu\gpu_device.cc:916] 0:   Y
I c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\core\common_runtime\gpu\gpu_device.cc:975] Creating TensorFlow device (/gpu:0) -> (device: 0, name: GeForce GTX 970, pci bus id: 0000:0f:00.0)
E c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\stream_executor\cuda\cuda_blas.cc:372] failed to create cublas handle: CUBLAS_STATUS_ALLOC_FAILED
W c:\tf_jenkins\home\workspace\release-win\device\gpu\os\windows\tensorflow\stream_executor\stream.cc:1390] attempting to perform BLAS operation using StreamExecutor without BLAS support
Traceback (most recent call last):
  File "C:\Users\Anonymous\AppData\Local\Programs\Python\Python35\lib\site-packages\tensorflow\python\client\session.py", line 1021, in _do_call
    return fn(*args)
  File "C:\Users\Anonymous\AppData\Local\Programs\Python\Python35\lib\site-packages\tensorflow\python\client\session.py", line 1003, in _run_fn
    status, run_metadata)
  File "C:\Users\Anonymous\AppData\Local\Programs\Python\Python35\lib\contextlib.py", line 66, in __exit__
    next(self.gen)
  File "C:\Users\Anonymous\AppData\Local\Programs\Python\Python35\lib\site-packages\tensorflow\python\framework\errors_impl.py", line 469, in raise_exception_on_not_ok_status
    pywrap_tensorflow.TF_GetCode(status))
tensorflow.python.framework.errors_impl.InternalError: Blas SGEMM launch failed : a.shape=(100, 784), b.shape=(784, 256), m=100, n=256, k=784
         [[Node: MatMul = MatMul[T=DT_FLOAT, transpose_a=false, transpose_b=false, _device="/job:localhost/replica:0/task:0/gpu:0"](_recv_Placeholder_0/_7, Variable/read)]]
         [[Node: Mean/_15 = _Recv[client_terminated=false, recv_device="/job:localhost/replica:0/task:0/cpu:0", send_device="/job:localhost/replica:0/task:0/gpu:0", send_device_incarnation=1, tensor_name="edge_35_Mean", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/cpu:0"]()]]

Answer 1

对于 TensorFlow 2.2，当遇到 CUBLAS_STATUS_ALLOC_FAILED 问题时，其他答案都不起作用。在https://www.tensorflow.org/guide/gpu上找到了解决方案:

import tensorflow as tf
gpus = tf.config.experimental.list_physical_devices('GPU')
if gpus:
    try:
        # Currently, memory growth needs to be the same across GPUs
        for gpu in gpus:
            tf.config.experimental.set_memory_growth(gpu, True)
        logical_gpus = tf.config.experimental.list_logical_devices('GPU')
        print(len(gpus), "Physical GPUs,", len(logical_gpus), "Logical GPUs")
    except RuntimeError as e:
        # Memory growth must be set before GPUs have been initialized
        print(e)

我在进行任何进一步计算之前运行了此代码，发现之前产生 CUBLAS 错误的相同代码现在在同一 session 中工作。上面的示例代码是一个具体示例，它设置了多个物理 GPU 之间的内存增长，但它也解决了内存扩展问题。