gemma.py

from openai import OpenAI

# initialize the client but point it to TGI
client = OpenAI(
  base_url="https://api-inference.huggingface.co/v1",
  api_key="hf_xxx" # Replace with your token
) 

chat_completion = client.chat.completions.create(
    model="google/gemma-7b-it",
    messages=[
        {"role": "user", "content": "Why is open-source software important?"},
    ],
    stream=True,
    max_tokens=500
)

# iterate and print stream
for message in chat_completion:
    print(message.choices[0].delta.content, end="")