Spaces:
Sleeping
Sleeping
| from flask import Flask, request, Response, jsonify | |
| from huggingface_hub import InferenceClient | |
| from flask_cors import CORS | |
| import json | |
| client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1") | |
| summary_prompt = '<s>[INST]Ты ассистент. Отвечаешь на русском языке. Сформируй краткое изложение следующего текста: {}[/INST]' | |
| def format_prompt(message, history): | |
| prompt = "<s>" | |
| for user_prompt, bot_response in history: | |
| prompt += f"[INST] {user_prompt} [/INST]" | |
| prompt += f" {bot_response}</s> " | |
| prompt += f"[INST] {message} [/INST]" | |
| return prompt | |
| def split_text(text): | |
| max_chars = 3500 | |
| sentences = text.split('.') | |
| lines = [] | |
| for sentence in sentences: | |
| lines.extend(sentence.split('\n')) | |
| result = [] | |
| current_chunk = '' | |
| for line in lines: | |
| if len(current_chunk) + len(line) < max_chars: | |
| current_chunk += line + '.' | |
| else: | |
| result.append(current_chunk.strip()) | |
| current_chunk = line + '.' | |
| if current_chunk: | |
| result.append(current_chunk.strip()) | |
| return result | |
| def generate( | |
| prompt, history=[], temperature=0, max_new_tokens=2000, top_p=0.95, repetition_penalty=1.0, | |
| ): | |
| temperature = float(temperature) | |
| if temperature < 1e-2: | |
| temperature = 1e-2 | |
| top_p = float(top_p) | |
| generate_kwargs = dict( | |
| temperature=temperature, | |
| max_new_tokens=max_new_tokens, | |
| top_p=top_p, | |
| repetition_penalty=repetition_penalty, | |
| do_sample=True, | |
| seed=42, | |
| ) | |
| #formatted_prompt = format_prompt(prompt, history) | |
| #stream = client.text_generation(prompt, **generate_kwargs, stream=True, details=False, return_full_text=False) | |
| response = client.text_generation(prompt, **generate_kwargs, stream=False, details=False, return_full_text=False) | |
| print(response) | |
| return response | |
| #output = "" | |
| #for response in stream: | |
| # yield response.token.text.encode('utf-8') | |
| app = Flask(__name__) | |
| CORS(app) | |
| def health(): | |
| return jsonify({"status": "ok"}) | |
| def completion_route(): | |
| data = request.get_json() | |
| prompt = data.get('prompt', '') | |
| #truncated_prompt = prompt[:32768] | |
| return Response(generate(prompt[:52768]), content_type='text/plain; charset=utf-8', status=200, direct_passthrough=True) | |
| def getsummary_route(): | |
| data = request.get_json() | |
| text = data.get('text', '') | |
| pages = split_text(text) | |
| result = '' | |
| for page in pages: | |
| summary_prompt = f'''<s>[INST]выпиши из текста в виде трех списков 1.какие вопросы, темы обсуждались, 2. какие проблемы были озвучены 3. какие предложения были сформулированы: {page}[/INST]''' | |
| response = generate(summary_prompt[:52000]) | |
| result = result + '\n'+response | |
| return jsonify({'result': result}) | |
| def cleantext_route(): | |
| data = request.get_json() | |
| text = data.get('text', '') | |
| summary_prompt = f'''<s>[INST]Ты ассистент. Отвечаешь на русском языке. Сформируй краткое изложение следующего текста: {text}[/INST]''' | |
| response = generate(summary_prompt[:52000]) | |
| return jsonify({'result': response}) | |
| def getfollowup_route(): | |
| data = request.get_json() | |
| text = data.get('text', '') | |
| summary_prompt = f'''<s>[INST]Ты ассистент. Отвечаешь на русском языке. Сформируй краткое изложение следующего текста: {text}[/INST]''' | |
| response = generate(summary_prompt[:52000]) | |
| return jsonify({'result': response}) | |
| def getagenda_route(): | |
| data = request.get_json() | |
| text = data.get('text', '') | |
| summary_prompt = f'''<s>[INST]Ты ассистент. Отвечаешь на русском языке. Сформируй краткое изложение следующего текста: {text}[/INST]''' | |
| response = generate(summary_prompt[:52000]) | |
| return jsonify({'result': response}) | |
| if __name__ == '__main__': | |
| app.run(debug=False, host='0.0.0.0', port=7860) |