W2 lab - testing fine tuned model from section 2.2

Course Q&A Generative AI with Large Language Models
1

Could be beneficial to add comments or examples to the notebook on how to actually use the trained model from section 2.2 - Fine-Tune the Model with the Preprocessed Dataset. Even this reduced example shows some improvements in the ROUGE metrics in section 2.4 - Evaluate the Model Quantitatively (with ROUGE Metric).

Not sure if this is the best way since this is the first time using these libraries, but I did this

trainer.train()
# add these to save the fine tuned model
trainer.save_model()
trainer.save_state()

load self tuned model instead of the checkpoint

# instruct_model = AutoModelForSeq2SeqLM.from_pretrained("./flan-dialogue-summary-checkpoint", torch_dtype=torch.bfloat16)
instruct_model = AutoModelForSeq2SeqLM.from_pretrained(output_dir, torch_dtype=torch.bfloat16)
2

Thanks for the feedback, Tom! Another learner pointed out something similar. We’ll consider this!

3

for this section I receive the following error

trainer.train()
’ ValueError: You have to specify either input_ids or inputs_embeds’
Any