FAQ

Q: What Python versions does the NPU compiler support?
A: Currently, it supports Python 2.7, Python 3.6, and Python 3.7.

Q: What TensorFlow versions does the NPU compiler support?
A: It supports all versions of TensorFlow between 1.4 and 1.15.

Q: What PyTorch versions does the NPU compiler support?
A: It supports all versions of PyTorch between 1.10 and 1.13, but it requires Python 3.7.

Q: When compiling a TensorFlow model, I encounter the error Some op types are not supported!. What should I do?
A: First, check if there are unnecessary ops for deployment, such as RandomUniform. If so, remove them. Otherwise, consult with Guoxin engineers to see if the unsupported OPs can be added.

Q: When compiling a PyTorch model, I encounter the error NPU Compiler don't support type xxx operator!. What should I do?
A: Consult with Guoxin engineers to see if the unsupported OPs can be added.

Q: Does the NPU compiler support dynamic RNN models?
A: No, it does not. You need to modify the model structure using a for loop.

Q: If there is an unclear error during compilation, what should I do?
A: Update the NPU compiler to the latest version. If the problem persists, consult with Guoxin engineers.