melodyguan
NCHW does not support when tring to run cifar10_micro_search
HI Guys:
when tring to run cifar10_micro_search, I always met the below error, I tried to switch to official tensorflow docker image, which should contain all op support, but it keeps same result.
Anyone can tell me how to do or share me the correct running env?
Thanks!
Best wishes!
Stanley
nvalidArgumentError (see above for traceback): Default AvgPoolingOp only supports NHWC on device type CPU [[Node: child/layer_0/cell_0/x/avg_pool/average_pooling2d/AvgPool = AvgPoolT=DT_FLOAT, data_format="NCHW", ksize=[1, 1, 3, 3], padding="SAME", strides=[1, 1, 1, 1], _device="/job:localhost/replica:0/task:0/device:CPU:0"]]
My cuda is 9.0. I changed tf.device("/device:GPU:0") from CPU and I used 'NCHW'. Then, uninstall and reinstall the following packages worked to me.
- tensorflow==1.8.0
- tensorflow-gpu==1.8.0 -tensorflow-tensorboard==1.5.1
same error happens bellow:
tensorflow.python.framework.errors_impl.InvalidArgumentError: Default AvgPoolingOp only supports NHWC on device type CPU [[Node: child/layer_0/cell_0/x/avg_pool/average_pooling2d/AvgPool = AvgPoolT=DT_FLOAT, data_format="NCHW", ksize=[1, 1, 3, 3], padding="SAME", strides=[1, 1, 1, 1], _device="/job:localhost/replica:0/task:0/device:CPU:0"]]
environment: tensorflow: 1.8 python: 3.6.5
I have solved this problem with the environment: tensorflow-gpu==1.13.1 python ==2.7
I have tried with different tensorflow version, and set allow_soft_placement=False, however still get stuck here.
- tensorflow-gpu==1.13.1, 1.14.0, 1.15.0
- python == 3.6.12
- cuda = 10.0
Any idea and suggestion would be welcome!
