README.md

metadata

language:
  - en
tags:
  - openvino

facebook/opt-13b

This is the facebook/opt-13b model converted to OpenVINO with INT8 weights compression for accelerated inference.

An example of how to do inference on this model:

from optimum.intel import OVModelForCausalLM
from transformers import AutoTokenizer, pipeline

# model_id should be set to either a local directory or a model available on the HuggingFace hub.
model_id = "helenai/facebook-opt-13b-ov"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = OVModelForCausalLM.from_pretrained(model_id)
pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)
result = pipe("hello world")
print(result)