- 신규로 표시
- 북마크
- 구독
- 소거
- RSS 피드 구독
- 강조
- 인쇄
- 부적절한 컨텐트 신고
Hi ,everyone! I encountered a problem. The following tensorflow code are from my training cnn code of mnist data.
num_input = 784
num_classes = 10
dropout = 0.75
X = tf.placeholder("float",[None,num_input],name="input")
Y = tf.placeholder("float",[256,num_classes],name="output")
keep_prob = tf.placeholder("float")
but when I use command mvNCCompile, mvNCCompile ckpt_555/model.ckpt.meta -s 12 -in=input -on=output -is 28 28 -o mnist_model.graph
there are some errors:
InvalidArgumentError (see above for traceback): Shape [-1,10] has negative dimensions
[[Node: output = Placeholder[dtype=DT_FLOAT, shape=[?,10], _device="/job:localhost/replica:0/task:0/cpu:0"]()]]
Can aynone help to tell me how to solve this problem.
링크가 복사됨
- 신규로 표시
- 북마크
- 구독
- 소거
- RSS 피드 구독
- 강조
- 인쇄
- 부적절한 컨텐트 신고
@guohua24 Please try the solution I mentioned at https://ncsforum.movidius.com/discussion/comment/1557/#Comment_1557 and let me know if this solves your issue.
- 신규로 표시
- 북마크
- 구독
- 소거
- RSS 피드 구독
- 강조
- 인쇄
- 부적절한 컨텐트 신고
@Tome_at_Intel
Please help me to solve the following problem.
when I use the following command: mvNCCompile tcs_inference.meta -w tcs_inference -s 12 -in input -on output -o tsc_inference.graph
the output is: mvNCCompile v02.00, Copyright @ Movidius Ltd 2016
/usr/local/lib/python3.5/dist-packages/tensorflow/python/util/tf_inspect.py:45: DeprecationWarning: inspect.getargspec() is deprecated, use inspect.signature() instead
if d.decorator_argspec is not None), _inspect.getargspec(target))
/usr/local/bin/ncsdk/Controllers/FileIO.py:52: UserWarning: You are using a large type. Consider reducing your data sizes for best performance
"Consider reducing your data sizes for best performance\033[0m")
- 신규로 표시
- 북마크
- 구독
- 소거
- RSS 피드 구독
- 강조
- 인쇄
- 부적절한 컨텐트 신고
@guohua24 If you are concerned about the warning You are using a large type. Consider reducing your data sizes for best performance
, I want you to know that it is an outdated warning and should be removed soon. This message shows up if you are using 32 bit data in your model, but since the weights for the model are always converted to fp16, you should not be worried about any performance issues.
