Spaces:
Running
Running
| import torch | |
| import gradio as gr | |
| import spaces | |
| from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer | |
| import os | |
| from threading import Thread | |
| import random | |
| from datasets import load_dataset | |
| import numpy as np | |
| from sklearn.feature_extraction.text import TfidfVectorizer | |
| import pandas as pd | |
| from typing import List, Tuple | |
| import json | |
| from datetime import datetime | |
| # GPU ๋ฉ๋ชจ๋ฆฌ ๊ด๋ฆฌ | |
| torch.cuda.empty_cache() | |
| # ํ๊ฒฝ ๋ณ์ ์ค์ | |
| HF_TOKEN = os.environ.get("HF_TOKEN", None) | |
| MODEL_ID = "CohereForAI/c4ai-command-r7b-12-2024" | |
| MODELS = os.environ.get("MODELS") | |
| MODEL_NAME = MODEL_ID.split("/")[-1] | |
| # ๋ชจ๋ธ๊ณผ ํ ํฌ๋์ด์ ๋ก๋ | |
| model = AutoModelForCausalLM.from_pretrained( | |
| MODEL_ID, | |
| torch_dtype=torch.bfloat16, | |
| device_map="auto", | |
| ) | |
| tokenizer = AutoTokenizer.from_pretrained(MODEL_ID) | |
| # ์ํคํผ๋์ ๋ฐ์ดํฐ์ ๋ก๋ | |
| wiki_dataset = load_dataset("lcw99/wikipedia-korean-20240501-1million-qna") | |
| print("Wikipedia dataset loaded:", wiki_dataset) | |
| # TF-IDF ๋ฒกํฐ๋ผ์ด์ ์ด๊ธฐํ ๋ฐ ํ์ต | |
| print("TF-IDF ๋ฒกํฐํ ์์...") | |
| questions = wiki_dataset['train']['question'][:10000] # ์ฒ์ 10000๊ฐ๋ง ์ฌ์ฉ | |
| vectorizer = TfidfVectorizer(max_features=1000) | |
| question_vectors = vectorizer.fit_transform(questions) | |
| print("TF-IDF ๋ฒกํฐํ ์๋ฃ") | |
| class ChatHistory: | |
| def __init__(self): | |
| self.history = [] | |
| self.history_file = "/tmp/chat_history.json" | |
| self.load_history() | |
| def add_conversation(self, user_msg: str, assistant_msg: str): | |
| conversation = { | |
| "timestamp": datetime.now().isoformat(), | |
| "messages": [ | |
| {"role": "user", "content": user_msg}, | |
| {"role": "assistant", "content": assistant_msg} | |
| ] | |
| } | |
| self.history.append(conversation) | |
| self.save_history() | |
| def format_for_display(self): | |
| formatted = [] | |
| for conv in self.history: | |
| formatted.append([ | |
| conv["messages"][0]["content"], | |
| conv["messages"][1]["content"] | |
| ]) | |
| return formatted | |
| def get_messages_for_api(self): | |
| messages = [] | |
| for conv in self.history: | |
| messages.extend([ | |
| {"role": "user", "content": conv["messages"][0]["content"]}, | |
| {"role": "assistant", "content": conv["messages"][1]["content"]} | |
| ]) | |
| return messages | |
| def clear_history(self): | |
| self.history = [] | |
| self.save_history() | |
| def save_history(self): | |
| try: | |
| with open(self.history_file, 'w', encoding='utf-8') as f: | |
| json.dump(self.history, f, ensure_ascii=False, indent=2) | |
| except Exception as e: | |
| print(f"ํ์คํ ๋ฆฌ ์ ์ฅ ์คํจ: {e}") | |
| def load_history(self): | |
| try: | |
| if os.path.exists(self.history_file): | |
| with open(self.history_file, 'r', encoding='utf-8') as f: | |
| self.history = json.load(f) | |
| except Exception as e: | |
| print(f"ํ์คํ ๋ฆฌ ๋ก๋ ์คํจ: {e}") | |
| self.history = [] | |
| # ์ ์ญ ChatHistory ์ธ์คํด์ค ์์ฑ | |
| chat_history = ChatHistory() | |
| def find_relevant_context(query, top_k=3): | |
| # ์ฟผ๋ฆฌ ๋ฒกํฐํ | |
| query_vector = vectorizer.transform([query]) | |
| # ์ฝ์ฌ์ธ ์ ์ฌ๋ ๊ณ์ฐ | |
| similarities = (query_vector * question_vectors.T).toarray()[0] | |
| # ๊ฐ์ฅ ์ ์ฌํ ์ง๋ฌธ๋ค์ ์ธ๋ฑ์ค | |
| top_indices = np.argsort(similarities)[-top_k:][::-1] | |
| # ๊ด๋ จ ์ปจํ ์คํธ ์ถ์ถ | |
| relevant_contexts = [] | |
| for idx in top_indices: | |
| if similarities[idx] > 0: | |
| relevant_contexts.append({ | |
| 'question': questions[idx], | |
| 'answer': wiki_dataset['train']['answer'][idx], | |
| 'similarity': similarities[idx] | |
| }) | |
| return relevant_contexts | |
| def analyze_file_content(content, file_type): | |
| """Analyze file content and return structural summary""" | |
| if file_type in ['parquet', 'csv']: | |
| try: | |
| lines = content.split('\n') | |
| header = lines[0] | |
| columns = header.count('|') - 1 | |
| rows = len(lines) - 3 | |
| return f"๐ ๋ฐ์ดํฐ์ ๊ตฌ์กฐ: {columns}๊ฐ ์ปฌ๋ผ, {rows}๊ฐ ๋ฐ์ดํฐ" | |
| except: | |
| return "โ ๋ฐ์ดํฐ์ ๊ตฌ์กฐ ๋ถ์ ์คํจ" | |
| lines = content.split('\n') | |
| total_lines = len(lines) | |
| non_empty_lines = len([line for line in lines if line.strip()]) | |
| if any(keyword in content.lower() for keyword in ['def ', 'class ', 'import ', 'function']): | |
| functions = len([line for line in lines if 'def ' in line]) | |
| classes = len([line for line in lines if 'class ' in line]) | |
| imports = len([line for line in lines if 'import ' in line or 'from ' in line]) | |
| return f"๐ป ์ฝ๋ ๊ตฌ์กฐ: {total_lines}์ค (ํจ์: {functions}, ํด๋์ค: {classes}, ์ํฌํธ: {imports})" | |
| paragraphs = content.count('\n\n') + 1 | |
| words = len(content.split()) | |
| return f"๐ ๋ฌธ์ ๊ตฌ์กฐ: {total_lines}์ค, {paragraphs}๋จ๋ฝ, ์ฝ {words}๋จ์ด" | |
| def read_uploaded_file(file): | |
| if file is None: | |
| return "", "" | |
| try: | |
| file_ext = os.path.splitext(file.name)[1].lower() | |
| if file_ext == '.parquet': | |
| df = pd.read_parquet(file.name, engine='pyarrow') | |
| content = df.head(10).to_markdown(index=False) | |
| return content, "parquet" | |
| elif file_ext == '.csv': | |
| encodings = ['utf-8', 'cp949', 'euc-kr', 'latin1'] | |
| for encoding in encodings: | |
| try: | |
| df = pd.read_csv(file.name, encoding=encoding) | |
| content = f"๐ ๋ฐ์ดํฐ ๋ฏธ๋ฆฌ๋ณด๊ธฐ:\n{df.head(10).to_markdown(index=False)}\n\n" | |
| content += f"\n๐ ๋ฐ์ดํฐ ์ ๋ณด:\n" | |
| content += f"- ์ ์ฒด ํ ์: {len(df)}\n" | |
| content += f"- ์ ์ฒด ์ด ์: {len(df.columns)}\n" | |
| content += f"- ์ปฌ๋ผ ๋ชฉ๋ก: {', '.join(df.columns)}\n" | |
| content += f"\n๐ ์ปฌ๋ผ ๋ฐ์ดํฐ ํ์ :\n" | |
| for col, dtype in df.dtypes.items(): | |
| content += f"- {col}: {dtype}\n" | |
| null_counts = df.isnull().sum() | |
| if null_counts.any(): | |
| content += f"\nโ ๏ธ ๊ฒฐ์ธก์น:\n" | |
| for col, null_count in null_counts[null_counts > 0].items(): | |
| content += f"- {col}: {null_count}๊ฐ ๋๋ฝ\n" | |
| return content, "csv" | |
| except UnicodeDecodeError: | |
| continue | |
| raise UnicodeDecodeError(f"โ ์ง์๋๋ ์ธ์ฝ๋ฉ์ผ๋ก ํ์ผ์ ์ฝ์ ์ ์์ต๋๋ค ({', '.join(encodings)})") | |
| else: | |
| encodings = ['utf-8', 'cp949', 'euc-kr', 'latin1'] | |
| for encoding in encodings: | |
| try: | |
| with open(file.name, 'r', encoding=encoding) as f: | |
| content = f.read() | |
| return content, "text" | |
| except UnicodeDecodeError: | |
| continue | |
| raise UnicodeDecodeError(f"โ ์ง์๋๋ ์ธ์ฝ๋ฉ์ผ๋ก ํ์ผ์ ์ฝ์ ์ ์์ต๋๋ค ({', '.join(encodings)})") | |
| except Exception as e: | |
| return f"โ ํ์ผ ์ฝ๊ธฐ ์ค๋ฅ: {str(e)}", "error" | |
| def read_uploaded_file(file): | |
| if file is None: | |
| return "", "" | |
| try: | |
| file_ext = os.path.splitext(file.name)[1].lower() | |
| if file_ext == '.parquet': | |
| df = pd.read_parquet(file.name) | |
| content = f"๐ ๋ฐ์ดํฐ ๋ฏธ๋ฆฌ๋ณด๊ธฐ:\n{df.head(10).to_markdown(index=False)}\n\n" | |
| content += f"\n๐ ๋ฐ์ดํฐ ์ ๋ณด:\n" | |
| content += f"- ์ ์ฒด ํ ์: {len(df)}\n" | |
| content += f"- ์ ์ฒด ์ด ์: {len(df.columns)}\n" | |
| content += f"- ์ปฌ๋ผ ๋ชฉ๋ก: {', '.join(df.columns)}\n" | |
| return content, "parquet" | |
| elif file_ext == '.csv': | |
| encodings = ['utf-8', 'cp949', 'euc-kr', 'latin1'] | |
| for encoding in encodings: | |
| try: | |
| df = pd.read_csv(file.name, encoding=encoding) | |
| content = f"๐ ๋ฐ์ดํฐ ๋ฏธ๋ฆฌ๋ณด๊ธฐ:\n{df.head(10).to_markdown(index=False)}\n\n" | |
| content += f"\n๐ ๋ฐ์ดํฐ ์ ๋ณด:\n" | |
| content += f"- ์ ์ฒด ํ ์: {len(df)}\n" | |
| content += f"- ์ ์ฒด ์ด ์: {len(df.columns)}\n" | |
| content += f"- ์ปฌ๋ผ ๋ชฉ๋ก: {', '.join(df.columns)}\n" | |
| content += f"\n๐ ์ปฌ๋ผ ๋ฐ์ดํฐ ํ์ :\n" | |
| for col, dtype in df.dtypes.items(): | |
| content += f"- {col}: {dtype}\n" | |
| null_counts = df.isnull().sum() | |
| if null_counts.any(): | |
| content += f"\nโ ๏ธ ๊ฒฐ์ธก์น:\n" | |
| for col, null_count in null_counts[null_counts > 0].items(): | |
| content += f"- {col}: {null_count}๊ฐ ๋๋ฝ\n" | |
| return content, "csv" | |
| except UnicodeDecodeError: | |
| continue | |
| raise UnicodeDecodeError(f"์ง์๋๋ ์ธ์ฝ๋ฉ์ผ๋ก ํ์ผ์ ์ฝ์ ์ ์์ต๋๋ค ({', '.join(encodings)})") | |
| else: # ํ ์คํธ ํ์ผ | |
| encodings = ['utf-8', 'cp949', 'euc-kr', 'latin1'] | |
| for encoding in encodings: | |
| try: | |
| with open(file.name, 'r', encoding=encoding) as f: | |
| content = f.read() | |
| # ํ์ผ ๋ด์ฉ ๋ถ์ | |
| lines = content.split('\n') | |
| total_lines = len(lines) | |
| non_empty_lines = len([line for line in lines if line.strip()]) | |
| # ์ฝ๋ ํ์ผ ์ฌ๋ถ ํ์ธ | |
| is_code = any(keyword in content.lower() for keyword in ['def ', 'class ', 'import ', 'function']) | |
| if is_code: | |
| # ์ฝ๋ ํ์ผ ๋ถ์ | |
| functions = len([line for line in lines if 'def ' in line]) | |
| classes = len([line for line in lines if 'class ' in line]) | |
| imports = len([line for line in lines if 'import ' in line or 'from ' in line]) | |
| analysis = f"\n๐ ์ฝ๋ ๋ถ์:\n" | |
| analysis += f"- ์ ์ฒด ๋ผ์ธ ์: {total_lines}\n" | |
| analysis += f"- ํจ์ ์: {functions}\n" | |
| analysis += f"- ํด๋์ค ์: {classes}\n" | |
| analysis += f"- import ๋ฌธ ์: {imports}\n" | |
| else: | |
| # ์ผ๋ฐ ํ ์คํธ ํ์ผ ๋ถ์ | |
| words = len(content.split()) | |
| chars = len(content) | |
| analysis = f"\n๐ ํ ์คํธ ๋ถ์:\n" | |
| analysis += f"- ์ ์ฒด ๋ผ์ธ ์: {total_lines}\n" | |
| analysis += f"- ์ค์ ๋ด์ฉ์ด ์๋ ๋ผ์ธ ์: {non_empty_lines}\n" | |
| analysis += f"- ๋จ์ด ์: {words}\n" | |
| analysis += f"- ๋ฌธ์ ์: {chars}\n" | |
| return content + analysis, "text" | |
| except UnicodeDecodeError: | |
| continue | |
| raise UnicodeDecodeError(f"์ง์๋๋ ์ธ์ฝ๋ฉ์ผ๋ก ํ์ผ์ ์ฝ์ ์ ์์ต๋๋ค ({', '.join(encodings)})") | |
| except Exception as e: | |
| return f"ํ์ผ ์ฝ๊ธฐ ์ค๋ฅ: {str(e)}", "error" | |
| # ํ์ผ ์ ๋ก๋ ์ด๋ฒคํธ ํธ๋ค๋ง ์์ | |
| def init_msg(): | |
| return "ํ์ผ์ ๋ถ์ํ๊ณ ์์ต๋๋ค..." | |
| CSS = """ | |
| /* 3D ์คํ์ผ CSS */ | |
| :root { | |
| --primary-color: #2196f3; | |
| --secondary-color: #1976d2; | |
| --background-color: #f0f2f5; | |
| --card-background: #ffffff; | |
| --text-color: #333333; | |
| --shadow-color: rgba(0, 0, 0, 0.1); | |
| } | |
| body { | |
| background: linear-gradient(135deg, #f5f7fa 0%, #c3cfe2 100%); | |
| min-height: 100vh; | |
| font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif; | |
| } | |
| .container { | |
| transform-style: preserve-3d; | |
| perspective: 1000px; | |
| } | |
| .chatbot { | |
| background: var(--card-background); | |
| border-radius: 20px; | |
| box-shadow: | |
| 0 10px 20px var(--shadow-color), | |
| 0 6px 6px var(--shadow-color); | |
| transform: translateZ(0); | |
| transition: transform 0.3s ease; | |
| backdrop-filter: blur(10px); | |
| } | |
| .chatbot:hover { | |
| transform: translateZ(10px); | |
| } | |
| /* ๋ฉ์์ง ์ ๋ ฅ ์์ญ */ | |
| .input-area { | |
| background: var(--card-background); | |
| border-radius: 15px; | |
| padding: 15px; | |
| margin-top: 20px; | |
| box-shadow: | |
| 0 5px 15px var(--shadow-color), | |
| 0 3px 3px var(--shadow-color); | |
| transform: translateZ(0); | |
| transition: all 0.3s ease; | |
| display: flex; | |
| align-items: center; | |
| gap: 10px; | |
| } | |
| .input-area:hover { | |
| transform: translateZ(5px); | |
| } | |
| /* ๋ฒํผ ์คํ์ผ */ | |
| .custom-button { | |
| background: linear-gradient(145deg, var(--primary-color), var(--secondary-color)); | |
| color: white; | |
| border: none; | |
| border-radius: 10px; | |
| padding: 10px 20px; | |
| font-weight: 600; | |
| cursor: pointer; | |
| transform: translateZ(0); | |
| transition: all 0.3s ease; | |
| box-shadow: | |
| 0 4px 6px var(--shadow-color), | |
| 0 1px 3px var(--shadow-color); | |
| } | |
| .custom-button:hover { | |
| transform: translateZ(5px) translateY(-2px); | |
| box-shadow: | |
| 0 7px 14px var(--shadow-color), | |
| 0 3px 6px var(--shadow-color); | |
| } | |
| /* ํ์ผ ์ ๋ก๋ ๋ฒํผ */ | |
| .file-upload-icon { | |
| background: linear-gradient(145deg, #64b5f6, #42a5f5); | |
| color: white; | |
| border-radius: 8px; | |
| font-size: 2em; | |
| cursor: pointer; | |
| display: flex; | |
| align-items: center; | |
| justify-content: center; | |
| height: 70px; | |
| width: 70px; | |
| transition: all 0.3s ease; | |
| box-shadow: 0 2px 5px rgba(0,0,0,0.1); | |
| } | |
| .file-upload-icon:hover { | |
| transform: translateY(-2px); | |
| box-shadow: 0 4px 8px rgba(0,0,0,0.2); | |
| } | |
| /* ํ์ผ ์ ๋ก๋ ๋ฒํผ ๋ด๋ถ ์์ ์คํ์ผ๋ง */ | |
| .file-upload-icon > .wrap { | |
| display: flex !important; | |
| align-items: center; | |
| justify-content: center; | |
| width: 100%; | |
| height: 100%; | |
| } | |
| .file-upload-icon > .wrap > p { | |
| display: none !important; | |
| } | |
| .file-upload-icon > .wrap::before { | |
| content: "๐"; | |
| font-size: 2em; | |
| display: block; | |
| } | |
| /* ๋ฉ์์ง ์คํ์ผ */ | |
| .message { | |
| background: var(--card-background); | |
| border-radius: 15px; | |
| padding: 15px; | |
| margin: 10px 0; | |
| box-shadow: | |
| 0 4px 6px var(--shadow-color), | |
| 0 1px 3px var(--shadow-color); | |
| transform: translateZ(0); | |
| transition: all 0.3s ease; | |
| } | |
| .message:hover { | |
| transform: translateZ(5px); | |
| } | |
| .chat-container { | |
| height: 600px !important; | |
| margin-bottom: 10px; | |
| } | |
| .input-container { | |
| height: 70px !important; | |
| display: flex; | |
| align-items: center; | |
| gap: 10px; | |
| margin-top: 5px; | |
| } | |
| .input-textbox { | |
| height: 70px !important; | |
| border-radius: 8px !important; | |
| font-size: 1.1em !important; | |
| padding: 10px 15px !important; | |
| display: flex !important; | |
| align-items: flex-start !important; /* ํ ์คํธ ์ ๋ ฅ ์์น๋ฅผ ์๋ก ์กฐ์ */ | |
| } | |
| .input-textbox textarea { | |
| padding-top: 5px !important; /* ํ ์คํธ ์๋จ ์ฌ๋ฐฑ ์กฐ์ */ | |
| } | |
| .send-button { | |
| height: 70px !important; | |
| min-width: 70px !important; | |
| font-size: 1.1em !important; | |
| } | |
| /* ์ค์ ํจ๋ ๊ธฐ๋ณธ ์คํ์ผ */ | |
| .settings-panel { | |
| padding: 20px; | |
| margin-top: 20px; | |
| } | |
| """ | |
| def stream_chat(message: str, history: list, uploaded_file, temperature: float, max_new_tokens: int, top_p: float, top_k: int, penalty: float): | |
| try: | |
| print(f'message is - {message}') | |
| print(f'history is - {history}') | |
| # ํ์ผ ์ ๋ก๋ ์ฒ๋ฆฌ | |
| file_context = "" | |
| if uploaded_file and message == "ํ์ผ์ ๋ถ์ํ๊ณ ์์ต๋๋ค...": | |
| try: | |
| content, file_type = read_uploaded_file(uploaded_file) | |
| if content: | |
| file_analysis = analyze_file_content(content, file_type) | |
| file_context = f"\n\n๐ ํ์ผ ๋ถ์ ๊ฒฐ๊ณผ:\n{file_analysis}\n\nํ์ผ ๋ด์ฉ:\n```\n{content}\n```" | |
| message = "์ ๋ก๋๋ ํ์ผ์ ๋ถ์ํด์ฃผ์ธ์." | |
| except Exception as e: | |
| print(f"ํ์ผ ๋ถ์ ์ค๋ฅ: {str(e)}") | |
| file_context = f"\n\nโ ํ์ผ ๋ถ์ ์ค ์ค๋ฅ๊ฐ ๋ฐ์ํ์ต๋๋ค: {str(e)}" | |
| # ๊ด๋ จ ์ปจํ ์คํธ ์ฐพ๊ธฐ | |
| try: | |
| relevant_contexts = find_relevant_context(message) | |
| wiki_context = "\n\n๊ด๋ จ ์ํคํผ๋์ ์ ๋ณด:\n" | |
| for ctx in relevant_contexts: | |
| wiki_context += f"Q: {ctx['question']}\nA: {ctx['answer']}\n์ ์ฌ๋: {ctx['similarity']:.3f}\n\n" | |
| except Exception as e: | |
| print(f"์ปจํ ์คํธ ๊ฒ์ ์ค๋ฅ: {str(e)}") | |
| wiki_context = "" | |
| # ๋ํ ํ์คํ ๋ฆฌ ๊ตฌ์ฑ | |
| conversation = [] | |
| for prompt, answer in history: | |
| conversation.extend([ | |
| {"role": "user", "content": prompt}, | |
| {"role": "assistant", "content": answer} | |
| ]) | |
| # ์ต์ข ํ๋กฌํํธ ๊ตฌ์ฑ | |
| final_message = file_context + wiki_context + "\nํ์ฌ ์ง๋ฌธ: " + message | |
| conversation.append({"role": "user", "content": final_message}) | |
| # ํ ํฌ๋์ด์ ์ค์ | |
| input_ids = tokenizer.apply_chat_template(conversation, tokenize=False, add_generation_prompt=True) | |
| inputs = tokenizer(input_ids, return_tensors="pt").to(0) | |
| streamer = TextIteratorStreamer(tokenizer, timeout=10., skip_prompt=True, skip_special_tokens=True) | |
| generate_kwargs = dict( | |
| inputs, | |
| streamer=streamer, | |
| top_k=top_k, | |
| top_p=top_p, | |
| repetition_penalty=penalty, | |
| max_new_tokens=max_new_tokens, | |
| do_sample=True, | |
| temperature=temperature, | |
| eos_token_id=[255001], | |
| ) | |
| thread = Thread(target=model.generate, kwargs=generate_kwargs) | |
| thread.start() | |
| buffer = "" | |
| for new_text in streamer: | |
| buffer += new_text | |
| yield "", history + [[message, buffer]] | |
| except Exception as e: | |
| error_message = f"์ค๋ฅ๊ฐ ๋ฐ์ํ์ต๋๋ค: {str(e)}" | |
| print(f"Stream chat ์ค๋ฅ: {error_message}") | |
| yield "", history + [[message, error_message]] | |
| def create_demo(): | |
| with gr.Blocks(css=CSS) as demo: | |
| chatbot = gr.Chatbot( | |
| value=[], | |
| height=600, | |
| label="GiniGEN AI Assistant", | |
| elem_classes="chat-container" | |
| ) | |
| with gr.Row(elem_classes="input-container"): | |
| with gr.Column(scale=1, min_width=70): | |
| file_upload = gr.File( | |
| type="filepath", | |
| elem_classes="file-upload-icon", | |
| scale=1, | |
| container=True, | |
| interactive=True, | |
| show_label=False | |
| ) | |
| with gr.Column(scale=4): | |
| msg = gr.Textbox( | |
| show_label=False, | |
| placeholder="๋ฉ์์ง๋ฅผ ์ ๋ ฅํ์ธ์... ๐ญ", | |
| container=False, | |
| elem_classes="input-textbox", | |
| scale=1 | |
| ) | |
| with gr.Column(scale=1, min_width=70): | |
| send = gr.Button( | |
| "์ ์ก", | |
| elem_classes="send-button custom-button", | |
| scale=1 | |
| ) | |
| with gr.Accordion("๐ฎ ๊ณ ๊ธ ์ค์ ", open=False): | |
| with gr.Row(): | |
| with gr.Column(scale=1): | |
| temperature = gr.Slider( | |
| minimum=0, maximum=1, step=0.1, value=0.8, | |
| label="์ฐฝ์์ฑ ์์ค ๐จ" | |
| ) | |
| max_new_tokens = gr.Slider( | |
| minimum=128, maximum=8000, step=1, value=4000, | |
| label="์ต๋ ํ ํฐ ์ ๐" | |
| ) | |
| with gr.Column(scale=1): | |
| top_p = gr.Slider( | |
| minimum=0.0, maximum=1.0, step=0.1, value=0.8, | |
| label="๋ค์์ฑ ์กฐ์ ๐ฏ" | |
| ) | |
| top_k = gr.Slider( | |
| minimum=1, maximum=20, step=1, value=20, | |
| label="์ ํ ๋ฒ์ ๐" | |
| ) | |
| penalty = gr.Slider( | |
| minimum=0.0, maximum=2.0, step=0.1, value=1.0, | |
| label="๋ฐ๋ณต ์ต์ ๐" | |
| ) | |
| gr.Examples( | |
| examples=[ | |
| ["๋ค์ ์ฝ๋์ ๋ฌธ์ ์ ์ ์ฐพ์๋ด๊ณ ๊ฐ์ ๋ ๋ฒ์ ์ ์ ์ํด์ฃผ์ธ์:\ndef fibonacci(n):\n if n <= 1: return n\n return fibonacci(n-1) + fibonacci(n-2)"], | |
| ["๋ค์ ์์ด ๋ฌธ์ฅ์ ํ๊ตญ์ด๋ก ๋ฒ์ญํ๊ณ , ์ดํ์ ๋ฌธ๋ฒ์ ํน์ง์ ์ค๋ช ํด์ฃผ์ธ์: 'The implementation of artificial intelligence in healthcare has revolutionized patient care, yet it raises ethical concerns regarding privacy and decision-making autonomy.'"], | |
| ["์ฃผ์ด์ง ๋ฐ์ดํฐ๋ฅผ ๋ถ์ํ๊ณ ์ธ์ฌ์ดํธ๋ฅผ ๋์ถํด์ฃผ์ธ์:\n์ฐ๋๋ณ ๋งค์ถ์ก(์ต์)\n2019: 1200\n2020: 980\n2021: 1450\n2022: 2100\n2023: 1890"], | |
| ["๋ค์ ์๋๋ฆฌ์ค์ ๋ํ SWOT ๋ถ์์ ํด์ฃผ์ธ์: '์ ํต์ ์ธ ์คํ๋ผ์ธ ์์ ์ด ์จ๋ผ์ธ ํ๋ซํผ์ผ๋ก์ ์ ํ์ ๊ณ ๋ ค์ค์ ๋๋ค. ๋ ์๋ค์ ๋์งํธ ์ฝํ ์ธ ์๋น๊ฐ ์ฆ๊ฐํ๋ ์ํฉ์์ ๊ฒฝ์๋ ฅ์ ์ ์งํ๋ฉด์ ๊ธฐ์กด ๊ณ ๊ฐ์ธต๋ ์งํค๊ณ ์ถ์ต๋๋ค.'"], | |
| ["๋ค์ ์ํ ๋ฌธ์ ๋ฅผ ๋จ๊ณ๋ณ๋ก ์์ธํ ํ์ดํด์ฃผ์ธ์: 'ํ ์์ ๋์ด๊ฐ ๊ทธ ์์ ๋ด์ ํ๋ ์ ์ฌ๊ฐํ ๋์ด์ 2๋ฐฐ์ผ ๋, ์์ ๋ฐ์ง๋ฆ๊ณผ ์ ์ฌ๊ฐํ์ ํ ๋ณ์ ๊ธธ์ด์ ๊ด๊ณ๋ฅผ ๊ตฌํ์์ค.'"], | |
| ["๋ค์ SQL ์ฟผ๋ฆฌ๋ฅผ ์ต์ ํํ๊ณ ๊ฐ์ ์ ์ ์ค๋ช ํด์ฃผ์ธ์:\nSELECT * FROM orders o\nLEFT JOIN customers c ON o.customer_id = c.id\nWHERE YEAR(o.order_date) = 2023\nAND c.country = 'Korea'\nORDER BY o.order_date DESC;"], | |
| ["๋ค์ ๋ง์ผํ ์บ ํ์ธ์ ROI๋ฅผ ๋ถ์ํ๊ณ ๊ฐ์ ๋ฐฉ์์ ์ ์ํด์ฃผ์ธ์:\n์ด ๋น์ฉ: 5000๋ง์\n๋๋ฌ์ ์: 100๋ง๋ช \nํด๋ฆญ๋ฅ : 2.3%\n์ ํ์จ: 0.8%\nํ๊ท ๊ตฌ๋งค์ก: 35,000์"], | |
| ], | |
| inputs=msg | |
| ) | |
| # ์ด๋ฒคํธ ๋ฐ์ธ๋ฉ | |
| msg.submit( | |
| stream_chat, | |
| inputs=[msg, chatbot, file_upload, temperature, max_new_tokens, top_p, top_k, penalty], | |
| outputs=[msg, chatbot] | |
| ) | |
| send.click( | |
| stream_chat, | |
| inputs=[msg, chatbot, file_upload, temperature, max_new_tokens, top_p, top_k, penalty], | |
| outputs=[msg, chatbot] | |
| ) | |
| file_upload.change( | |
| fn=init_msg, | |
| outputs=msg, | |
| queue=False | |
| ).then( | |
| fn=stream_chat, | |
| inputs=[msg, chatbot, file_upload, temperature, max_new_tokens, top_p, top_k, penalty], | |
| outputs=[msg, chatbot], | |
| queue=True | |
| ) | |
| return demo | |
| if __name__ == "__main__": | |
| demo = create_demo() | |
| demo.launch() |