How to train the model on my data to get more accurate results.
#4
by
Mustafagamal
- opened
What are the practices for transfer learning and hyperparameter optimization, ensuring the model adapts effectively to the nuances of my data
Hi Mustafa,
You may directly apply transfer learning to the current architecture (see model.py in the Files and Versions). If your data is more complex, I advise you to add layers.
In order to create a dataset from your data, you must format the input as follows:
inputs = tokenizer(product1, product2, padding='max_length', truncation=True, max_length=350, return_tensors='pt')
And for the output, you may use binary classification indicating whether the products concateneated are the same product or not. You may also use floats to enhance the model behaviour for unsure predictions.
Hope this helps.