Open sreenivasmrpivot opened 6 months ago
As per the https://github.com/ShishirPatil/gorilla/blob/main/inference/README.md , based on the Llama agreement, we will have to merge the Gorilla weights with the base Llama weights using the https://github.com/ShishirPatil/gorilla/blob/main/inference/apply_delta.py.
As per Lllama 3 documentation here https://github.com/meta-llama/llama3, the weights are available in both native llama3 format and huggingface format. which should we be using for the apply_delta.py?
Would the Gorilla Delta weights work with Llama 3 native weights?
If at all we have all the above queries taken care of, should we be applying Gorilla delta on base Lllama model weights or Instruct based model weights?
+1 to this question
+1 to this question
Are there any plans to release Gorilla based on Llama 3? Is there any readme where we can find instructions on fine-tuning Llama 3 for Gorilla?