Closed Cloudac7 closed 3 weeks ago
When the type map is not given (which is a bug) and the data has different type maps, it is expected to raise an error, otherwise we don't know which type map should be used for the model. So #4031 should be the correct way to fix it. However, the error message should be improved. The current one is for developers but not users.
Bug summary
It seems that in 3.0.0b3, executing multitask training or finetune task would run into a RuntimeError, calling inconsistent type map, while the same case could run perfectly on code installed from 2024Q1 branch. The original
input.json
is uploaded, to identify the bug.And in #4031, a possible solution to this issue is addressed, but it is not the direct error raised.
DeePMD-kit Version
3.0.0b3
Backend and its version
PyTorch v2.0.0.post200, TensorFlow v2.14.0
How did you download the software?
Offline packages
Input Files, Running Commands, Error Log, etc.
input.json
Steps to Reproduce
Please run a multitask training using dataset from
Domains_Cluster
.Further Information, Files, and Links
No response