Huawei uses machine translation combined with human proofreading to translate this document to different languages in order to help you better understand the content of this document.
Note: Even the most advanced machine translation cannot match the quality of professional translators.
Huawei shall not bear any responsibility for translation accuracy and it is recommended that you refer to the English document (a link for which has been provided).
Conversion Failure Caused by the Default Value of input_shape
Symptom
Model conversion fails, and the log message "input x shape is empty" is printed, as shown in Figure 3-4.
Possible Cause
The possible cause is as follows:
The input of the original TensorFlow model has a dynamic shape, for example, ? x 299 x 299 x 3. The default value of input_shape is used during model conversion. As a result, an error is reported.
Solution
- Use the Netron tool to view the model and confirm the input shape of the model.
- For a dynamic shape model, the input_shape argument must be added during conversion. Command example:
atc --model=./resnetv2.pb --framework=3 --input_shape="input_tensor:8,299,299,3" --input_format="NHWC" --output=./resnetv2 --soc_version=Ascend310
