When I try to do some picture-classification work,using cpu I get accuracy 0.9up,but using GPU I can only get accuracy 0.0x

**System Information (please complete the following information):**
- Model Builder Version (available in Manage Extensions dialog): 17.17.0.2332602
- Visual Studio Version vs2022

**Describe the bug**
- On which step of the process did you run into an issue: after train the model
- Clear description of the problem: When I try to do some picture-classification work,using cpu I get accuracy 0.9up,but using GPU I can only get accuracy 0.0x

**To Reproduce**
Steps to reproduce the behavior:
1. Go to 'MLModel1.mbconfig'
2. Click on '环境'
3. Scroll down to '训练'
4. See error

**Expected behavior**
A clear and concise description of what you expected to happen.
the accuracy should be just as the cpu's
**Screenshots**
If applicable, add screenshots to help explain your problem.

**Additional context**
Add any other context about the problem here.
my gpu is rtx4060