The issue of model fine-tuning
#10
by
Danfoyin
- opened
Hello, we have tried fine-tuning methods, but found that the improvements are limited to individual time series data, while the overall performance of the fine-tuned model actually decreases. Do you have any suggestions for improving the generalization of the model in specific scenarios?
Additionally, when fine-tuning with certain specific datasets, we found that the model fails to make predictions (the outputs are NaN), but after checking the data, we didn’t notice anything unusual. Could you help us understand what might be causing this issue?
Thank you very much!