rayrayraykk
commented
9 months ago It appears that the model has not been successfully trained. Potential solutions, which are not exhaustive, include:
- Check Training Process: Analyze the training logs and metrics to see if the model was learning appropriately. Check for signs of overfitting or underfitting, and make sure the loss was decreasing over time.
- Hyperparameter Tuning: Adjust the learning rate, batch size, number of epochs, and other relevant hyperparameters. Sometimes a model outputting null values can be a sign that the learning rate is too high or too low.
- Input Formatting: Verify that the inputs during testing are formatted in the same way as during training. The model might be sensitive to specific formats or keywords that indicate the start and end of an input.
Hi，
Thanks for your code. I want to ask for some help.
After training on my own dataset using LLM, the model output is null during testing, how can I solve this problem?
The train dataset is IMDB dataset, 'instruction', 'input', and 'output' format, and the test is also given instructions and inputs, the I want the output to be positive or negative, but the output is often null.