Llama-2-Open-Source-LLM-CPU-Inference
Llama-2-Open-Source-LLM-CPU-Inference copied to clipboard
Any ideas what "Illegal Instruction" indicates?
- Using
llama-2-7b-chat.ggmlv3.q8_0.bin
downloaded today - I've setup everything running on a Dell PowerEdge 16-processor machine w/ 128GB of RAM.
- Pointed to a directory that only had 4 PDFs in it.
- Every question just using your standard
poetry run python main.py "question text?"
returns just one line:
"Illegal instruction"
Some questions take longer to return the same.
Any clues?
Maybe you are running on a VM environment, and cat /proc/cpuinfo | grep flags
has no avx
or certain required flag.
In such case, try to adjust VM CPU settings.