OpenVINO 集成

pip install -U-strategy eager "optimum[openvino,nncf]" langchain-huggingface --quiet

from langchain_huggingface import HuggingFacePipeline

ov_config = {"PERFORMANCE_HINT": "LATENCY", "NUM_STREAMS": "1", "CACHE_DIR": ""}

ov_llm = HuggingFacePipeline.from_model_id(
    model_id="gpt2",
    task="text-generation",
    backend="openvino",
    model_kwargs={"device": "CPU", "ov_config": ov_config},
    pipeline_kwargs={"max_new_tokens": 10},
)

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

model_id = "gpt2"
device = "CPU"
tokenizer = AutoTokenizer.from_pretrained(model_id)
ov_model = OVModelForCausalLM.from_pretrained(
    model_id, export=True, device=device, ov_config=ov_config
)
ov_pipe = pipeline(
    "text-generation", model=ov_model, tokenizer=tokenizer, max_new_tokens=10
)
ov_llm = HuggingFacePipeline(pipeline=ov_pipe)

from langchain_core.prompts import PromptTemplate

template = """Question: {question}

Answer: Let's think step by step."""
prompt = PromptTemplate.from_template(template)

chain = prompt | ov_llm

question = "What is electroencephalography?"

print(chain.invoke({"question": question}))

chain = prompt | ov_llm.bind(skip_prompt=True)

question = "What is electroencephalography?"

print(chain.invoke({"question": question}))

!optimum-cli export openvino --model gpt2 ov_model_dir

!optimum-cli export openvino --model gpt2  --weight-format int8 ov_model_dir # 用于 8 位量化

!optimum-cli export openvino --model gpt2  --weight-format int4 ov_model_dir # 用于 4 位量化

ov_llm = HuggingFacePipeline.from_model_id(
    model_id="ov_model_dir",
    task="text-generation",
    backend="openvino",
    model_kwargs={"device": "CPU", "ov_config": ov_config},
    pipeline_kwargs={"max_new_tokens": 10},
)

chain = prompt | ov_llm

question = "What is electroencephalography?"

print(chain.invoke({"question": question}))

ov_config = {
    "KV_CACHE_PRECISION": "u8",
    "DYNAMIC_QUANTIZATION_GROUP_SIZE": "32",
    "PERFORMANCE_HINT": "LATENCY",
    "NUM_STREAMS": "1",
    "CACHE_DIR": "",
}

generation_config = {"skip_prompt": True, "pipeline_kwargs": {"max_new_tokens": 100}}
chain = prompt | ov_llm.bind(**generation_config)

for chunk in chain.stream(question):
    print(chunk, end="", flush=True)