Model inference issue

[s5110851108]

sorry to bother you ,I have written an inference script that uses the model parameters saved during training for inference. While the script is able to perform inference, I found that the results differ from those during the TEST phase (using the same data). Upon checking the model architecture, I discovered that many important parameters (e.g., the parameters of the embedding layer) were not saved when the model was stored. Is this normal?How can i solve theese prolem?