Hello everyone,
I’m currently working on a machine learning project where I need to deploy a pre-trained BERT model for sentiment analysis. Here’s a brief overview of what I’ve done so far:
- Downloaded and Fine-Tuned the Model: I successfully downloaded a pre-trained BERT model from Hugging Face and fine-tuned it on a custom dataset for binary sentiment classification (positive/negative).
-
Compiled the Model Using TVM: After fine-tuning, I compiled the model using TVM, generating a shared object file (
libmodel.so) to optimize inference performance on my CPU.
Now, I’m at the stage where I need to test this compiled model to ensure it works as expected before deploying it in a production environment.
Questions:
- Is the testing approach I’m using correct for ensuring that the compiled model functions properly?
- Are there any specific considerations I should keep in mind regarding input dimensions or preprocessing?
- What steps can I take to further validate the model’s predictions?
- How can test the model with the real word application?