Fine-Tune Llama on main and auxiliary task

mike_plumbs

I am trying to fine-tune Llama model on two task at the same time, using hugging face library: Main task: Causal language model like the model was initially trained for A classification task based on the whole input sequence (recommend an article). For this task I am getting as a reference the LlamaForCausalLM class, overwriting init and forward functions . However, I want to combine the two tasks above into one process. The main problem is that...

Originally asked by Dimits on Stack Exchange (ai) (CC BY-SA 4.0).

uphilltech

Great question — answering as the UphillTech team:

After your forward pass, you can always just use the logits and calculate the loss yourself. In this case, you'd only use the loss on the last input token. Huggingface also has a LlamaForSequenceClassification class, which does exactly this.

Full answer by Alexander Wan: https://ai.stackexchange.com/a/41303 (CC BY-SA 4.0).

Need it done for your business? UphillTech.