elyza-jp-llama2-7b

elyza-jp-llama2-7b.py

import openai

openai.api_base = "https://elyza-jp-llama2-7b.lepton.run/api/v1"
openai.api_key = "YOUR_API_KEY"

B_INST, E_INST = "[INST]", "[/INST]"
B_SYS, E_SYS = "<<SYS>>\n", "\n<</SYS>>\n\n"
DEFAULT_SYSTEM_PROMPT = "あなたは誠実で優秀な日本人のアシスタントです。"
BOS_TOKEN = "<s>"
text = "クマが海辺に行ってアザラシと友達になり、最終的には家に帰るというプロットの短編小説を書いてください。"


prompt = "{bos_token}{b_inst} {system}{prompt} {e_inst} ".format(
    bos_token=BOS_TOKEN,
    b_inst=B_INST,
    system=f"{B_SYS}{DEFAULT_SYSTEM_PROMPT}{E_SYS}",
    prompt=text,
    e_inst=E_INST,
)

print("===============PROMPT===================")
print(prompt)
print("===============ANSWER===================")
completion = openai.Completion.create(
    model="elyza-jp-llama2-7b",  # can be "gpt-3.5-turbo" as well
    prompt=prompt,
    max_tokens=512,
    stream=True,
)

for comp in completion:
    text = comp["choices"][0]["text"]
    print(text, end="")
print("")