Spaces:
Sleeping
Sleeping
| import os | |
| import streamlit as st | |
| from langchain.llms import HuggingFaceHub | |
| from models import return_sum_models | |
| class LLM_Langchain(): | |
| def __init__(self): | |
| st.warning("Warning: input function needs cleaning and may take long to be processed") | |
| st.header('🦜 Code summarization with CodeT5-small') | |
| st.warning("I do sample subsets of CodeXGLUE to finetune !") | |
| self.API_KEY = os.environ["API_KEY"] | |
| languages = ["php", "java", "javascript", "python", "ruby"] | |
| model_parent = st.sidebar.selectbox( | |
| label = "Choose Language", | |
| options = languages, | |
| help="Choose languages", | |
| ) | |
| if model_parent is None: | |
| model_name_visibility = True | |
| else: | |
| model_name_visibility = False | |
| self.model_name = return_sum_models(model_parent) | |
| checkpoint = st.sidebar.selectbox( | |
| label = "Model used", | |
| options = [self.model_name], | |
| help="Model used to predict", | |
| ) | |
| self.max_new_tokens = st.sidebar.slider( | |
| label="Token Length", | |
| min_value=32, | |
| max_value=1024, | |
| step=32, | |
| value=120, | |
| help="Set the max tokens to get accurate results" | |
| ) | |
| self.top_k = st.sidebar.slider( | |
| label="top_k", | |
| min_value=1.0, | |
| max_value=50.0, | |
| step=1.0, | |
| value=30.0, | |
| help="Set the top_k" | |
| ) | |
| self.top_p = st.sidebar.slider( | |
| label="top_p", | |
| min_value=0.1, | |
| max_value=1.0, | |
| step=0.05, | |
| value=0.95, | |
| help="Set the top_p" | |
| ) | |
| self.model_kwargs = { | |
| "max_new_tokens": self.max_new_tokens, | |
| "top_k": self.top_k, | |
| "top_p": self.top_p | |
| } | |
| os.environ['HUGGINGFACEHUB_API_TOKEN'] = self.API_KEY | |
| def generate_response(self, input_text): | |
| llm = HuggingFaceHub( | |
| repo_id = self.model_name, | |
| model_kwargs = self.model_kwargs | |
| ) | |
| return llm(input_text) | |
| def form_data(self): | |
| # with st.form('my_form'): | |
| try: | |
| if not self.API_KEY.startswith('hf_'): | |
| st.warning('Please enter your API key!', icon='⚠') | |
| if "messages" not in st.session_state: | |
| st.session_state.messages = [] | |
| st.write(f"You are using {self.model_name} model") | |
| for message in st.session_state.messages: | |
| with st.chat_message(message.get('role')): | |
| st.write(message.get("content")) | |
| text = st.chat_input(disabled=False) | |
| if text: | |
| st.session_state.messages.append( | |
| { | |
| "role":"user", | |
| "content": text | |
| } | |
| ) | |
| with st.chat_message("user"): | |
| st.write(text) | |
| if text.lower() == "clear": | |
| del st.session_state.messages | |
| return | |
| result = self.generate_response(text) | |
| st.session_state.messages.append( | |
| { | |
| "role": "assistant", | |
| "content": result | |
| } | |
| ) | |
| with st.chat_message('assistant'): | |
| st.markdown(result) | |
| except Exception as e: | |
| st.error(e, icon="🚨") | |
| model = LLM_Langchain() | |
| model.form_data() | |