Error trying to submit LLaMA 2 base model

#131
by artidoro - opened

Hello, I tried submitting a model with meta-llama/Llama-2-70b-hf as base model and got the error "base model not found on hub". I suppose this is because it requires an authorization token. Is there a way around this?

Open LLM Leaderboard org
This comment has been hidden
Open LLM Leaderboard org

Hi @artidoro , What was the model you submitted ?

Open LLM Leaderboard org
edited Jul 27, 2023

If the model you submitted was as delta weights on top of meta-llama/Llama-2-70b-hf, yes, it is because accessing the meta-llama base models requires an authentication token.
Since the license allows it, could you merge the weights yourself and submit the merged model to the leaderboard? (As "Original" weights)

Thank you! That makes sense. Would be nice to be able to use adapters directly on top of llama 2 without having to merge. But this will work for me

clefourrier changed discussion status to closed

Sign up or log in to comment