KingZack/syncing-github-to-huggingface

1import streamlit as st2from huggingface_hub import InferenceClient3 4 5# MUST SET HF_TOKEN IN STREAMLIT SETTINGS IN HUGGINGFACE REPO SECRETS6HF_TOKEN = st.secrets["HF_TOKEN"]7 8 9# INIT THE INFERENCE CLIENT WITH YOUR HF TOKEN10client = InferenceClient(11    provider="hf-inference",12    api_key=HF_TOKEN,13)14 15# THIS IS JUST THE streamlit TEXT INPUT WIDGET16user_input = st.text_input(17    "Place your prompt here",18    "This is a placeholder",19    key="placeholder",20)21 22# THIS IS THE INFERENCE CLIENT CALL23completion = client.chat.completions.create(24    model="HuggingFaceH4/zephyr-7b-beta",25    messages=[26        {27            "role": "user",28            "content": user_input29        }30    ],31    max_tokens=512,32)33 34# THIS IS THE RESPONSE FROM THE INFERENCE CLIENT35ai_response = completion.choices[0].message.content36 37# THIS IS THE STREAMLIT TEXT OUTPUT WIDGET WITH THE RESPONSE FROM THE INFERENCE CLIENT38st.text(ai_response)39 40 41### WRONG WAY TO TRY AND LOAD MODELS::: 42# Load model directly43# from transformers import AutoTokenizer, AutoModelForCausalLM44 45# tokenizer = AutoTokenizer.from_pretrained("deepseek-ai/DeepSeek-Prover-V2-671B", trust_remote_code=True)46# model = AutoModelForCausalLM.from_pretrained("deepseek-ai/DeepSeek-Prover-V2-671B", trust_remote_code=True)