Llamacpp error

#1
by ML-master-123 - opened

llm = LlamaCpp(
model_path=r"/home/ubuntu/FAQ-LLM/Llama-3.1-8B-Lexi-Uncensored_V2_Q8.gguf",
temperature=0,
max_tokens=2000,
n_gpu_layers=-1,
n_batch=128,
top_p=1,
callback_manager=callback_manager,
verbose=True,
n_ctx=2048
)

Facing an issue while loading the model it is giving the error of
Error while loading LLM model: 1 validation error for LlamaCpp

I've also tried f16 model but getting same issue

@dj13 Did you ensure to update llama.cpp to the latest version? They have fixed rope issues with Llama 3.1 since a while back and it's not backwards compatible.

Orenguteng changed discussion status to closed

Does anyone know if Ooba uses the latest llama.cpp for this to work or not? Or do I need to stick with version one of this model and grab the backwards compatible one.

Orenguteng changed discussion status to open
Orenguteng changed discussion status to closed

Sign up or log in to comment