Question

运行以下代码时，出现错误：

E tensorflow/stream_executor/cuda/cuda_blas.cc:654] failed to run cuBLAS routine cublasSgemv_v2: CUBLAS_STATUS_EXECUTION_FAILED
Traceback (most recent call last):
File "modelAndLayer.py", line 16, in <module>
y_pred=model(X)
File "/home/cxsbg/anaconda3/envs/dl36/lib/python3.6/site-packages/tensorflow/python/keras/_impl/keras/engine/base_layer.py", line 314, in __call__
output = super(Layer, self).__call__(inputs, *args, **kwargs)
File "/home/cxsbg/anaconda3/envs/dl36/lib/python3.6/site-packages/tensorflow/python/layers/base.py", line 717, in __call__
outputs = self.call(inputs, *args, **kwargs)
File "modelAndLayer.py", line 10, in call
output=self.dense(input)
File "/home/cxsbg/anaconda3/envs/dl36/lib/python3.6/site-packages/tensorflow/python/keras/_impl/keras/engine/base_layer.py", line 314, in __call__
output = super(Layer, self).__call__(inputs, *args, **kwargs)
File "/home/cxsbg/anaconda3/envs/dl36/lib/python3.6/site-packages/tensorflow/python/layers/base.py", line 717, in __call__
outputs = self.call(inputs, *args, **kwargs)
File "/home/cxsbg/anaconda3/envs/dl36/lib/python3.6/site-packages/tensorflow/python/layers/core.py", line 163, in call
outputs = gen_math_ops.mat_mul(inputs, self.kernel)
File "/home/cxsbg/anaconda3/envs/dl36/lib/python3.6/site-packages/tensorflow/python/ops/gen_math_ops.py", line 4305, in mat_mul
_six.raise_from(_core._status_to_exception(e.code, message), None)
File "<string>", line 3, in raise_from
tensorflow.python.framework.errors_impl.InternalError: Blas GEMV launch failed:  m=3, n=2 [Op:MatMul]

我的GPU是 RTX2080 ，驱动程序是 v410 。 cuda为 v9.0 ，cudnn为 v7 。 tensorflow-gpu是 v1.8 （我对v1.8和v1.12都感到厌倦）。 python是 v3.6 （我在v3.6和v2.7上都尝试过）。系统为 Ubuntu 16.04 （我对win10也感到厌倦）。

该问题总是在tensorflow-gpu上发生，但在tensorflow cpu上有效。

代码在这里（一个简单的线性模型）：

import tensorflow as tf
tf.enable_eager_execution()
X=tf.constant([[1.,2.,3,],[4.,5.,6.]])
Y=tf.constant([[10.],[20.]])
class Linear(tf.keras.Model):
    def __init__(self):
        super().__init__()
        self.dense=tf.keras.layers.Dense(units=1,kernel_initializer=tf.zeros_initializer(),bias_initializer=tf.zeros_initializer())
    def call(self,input):
        output=self.dense(input)
        return output
model=Linear()
optimizer=tf.train.GradientDescentOptimizer(learning_rate=1e-3)
for i in range(1000):
    with tf.GradientTape() as tape:
        y_pred=model(X)
        loss=tf.reduce_mean(tf.square(y_pred-Y))
    grads=tape.gradient(loss,model.variables)
    optimizer.apply_gradients(zip(grads,model.variables))
print(model.variables)

Answer 1

我认为该错误是由tf.enable_eager_execution（）引起的，因为我对其进行了多次测试。感谢作者which-version-of-cuda-can-work-with-rtx-2080。当我使用cuda9.2时，该错误已修复。

Blas GEMV启动失败：m = 3，n = 2 [Op：MatMul]

1 个答案: