Closed Jacob-Langley closed 11 months ago
What version of llama-cpp-python
are you using?
Does the error still occur if you supply verbose=True
to LLM
?
llm = LLM(use_larger=True, n_gpu_layers=35, verbose=True)
verbose=True appears to have done the trick! -- (why?)
Due to what seems to be a bug in llama-cpp-python.
It appears to download wizardlm-13b-v1.2.Q4_K_M.gguf without issue, but then gives the below when I try to use LLM().
Same for the 7B
llm = LLM(use_larger=True, n_gpu_layers=35) Traceback (most recent call last):
Cell In[4], line 1 llm = LLM(use_larger=True, n_gpu_layers=35)
File ~\anaconda3\lib\site-packages\onprem\core.py:119 in init self.load_llm()
File ~\anaconda3\lib\site-packages\onprem\core.py:226 in load_llm self.llm = llm = LlamaCpp(
File ~\anaconda3\lib\site-packages\langchain\load\serializable.py:74 in init super().init(**kwargs)
File pydantic\main.py:341 in pydantic.main.BaseModel.init
ValidationError: 1 validation error for LlamaCpp root Could not load Llama model from path: C:\Users\jlangley\onprem_data\wizardlm-13b-v1.2.Q4_K_M.gguf. Received error fileno (type=value_error)