reflector/server/trials/title_summary/chat_llm.py

"""
This is an example code containing the bare essentials to load a chat
 LLM and infer from it using a predefined prompt. The purpose of this file
 is to show an example of inferring from a chat LLM which is required for
 banana.dev due to its design and platform limitations
"""

# The following logic was tested on the monadical-ml machine

import json

import torch
from transformers import (
    AutoModelForCausalLM,
    AutoTokenizer
)
from transformers.generation import GenerationConfig

# This can be passed via the environment variable or the params supplied
# when starting the program via banana.dev platform
MODEL_NAME = "lmsys/vicuna-13b-v1.5"

# Load the model in half precision, and less memory usage
model = AutoModelForCausalLM.from_pretrained(MODEL_NAME,
                                             low_cpu_mem_usage=True,
                                             torch_dtype=torch.bfloat16
                                             )

# Generation config
model.config.max_new_tokens = 300
gen_cfg = GenerationConfig.from_model_config(model.config)
gen_cfg.max_new_tokens = 300

# Load the tokenizer
tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)

# Move model to GPU
model = model.cuda()
print(f"Loading {MODEL_NAME} successful")

# Inputs
sample_chunks = [
        "You all just came off of your incredible Google Cloud next conference where you released a wide variety of functionality and features and new products across artisan television and also across the entire sort of cloud ecosystem . You want to just first by walking through , first start by walking through all the innovations that you sort of released and what you 're excited about when you come to Google Cloud ? Now our vision is super simple .  If you look at what smartphones did for a consumer , you know they took a computer and internet browser , a communication device , and a camera , and made it so that it 's in everybody 's pocket , so it really brought computation to every person . We feel that , you know , our , what we 're trying to do is take all the technological innovation that Google 's doing , but make it super simple so that everyone can consume it . And so that includes our global data center footprint , all the new types of hardware and large-scale systems we work on , the software that we 're making available for people to do high-scale computation , tools for data processing , tools for cybersecurity , processing , tools for cyber security , tools for machine learning , but make it so simple that everyone can use it . And every step that we do to simplify things for people , we think adoption can grow . And so that 's a lot of what we 've done these last three , four years , and we made a number of announcements that next in machine learning and AI in particular , you know , we look at our work as four elements , how we take our large-scale compute systems that were building for AI and how we make that available to everybody .  Second , what we 're doing with the software stacks and top of it , things like jacks and other things and how we 're making those available to everybody . Third is advances because different people have different levels of expertise . Some people say I need the hardware to build my own large language model or algorithm . Other people say , look , I really need to use a building block . You guys give me .  So , 30s we 've done a lot with AutoML and we announce new capability for image , video , and translation to make it available to everybody .  And then lastly , we 're also building completely packaged solutions for some areas and we announce some new stuff . ",
        " We 're joined next by Thomas Curian , CEO of Google Cloud , and Alexander Wang , CEO and founder of Scale AI . Thomas joined Google in November 2018 as the CEO of Google Cloud . Prior to Google , Thomas spent 22 years at Oracle , where most recently he was president of product development . Before that , Thomas worked at McKinsey as a business analyst and engagement manager . His nearly 30 years of experience have given him a deep knowledge of engineering enterprise relationships and leadership of large organizations . Thomas 's degrees include an MBA in administration and management from Stanford University , as an RJ Miller scholar and a BSEE in electrical engineering and computer science from Princeton University , where he graduated suma cum laude .  Thomas serves as a member of the Stanford graduate School of Business Advisory Council and Princeton University School of Engineering Advisory Council . Please welcome to the stage , Thomas Curian and Alexander Wang . This is a super exciting conversation . Thanks for being here , Thomas ."]

# Model Prompt template for current model
prompt = f"""
        ### Human:
        Create a JSON object as response.The JSON object must have 2 fields:
        i) title and ii) summary.For the title field,generate a short title
        for the given text. For the summary field, summarize the given text
        in three sentences.

        {sample_chunks[0]}

        ### Assistant:
        """

# Inference : Chat generation
input_ids = tokenizer.encode(prompt, return_tensors='pt').to(model.device)
output = model.generate(input_ids, generation_config=gen_cfg)

# Process output
response = tokenizer.decode(output[0].cpu(), skip_special_tokens=True)
response = response.split("### Assistant:\n")
print("TitleSummaryJsonResponse :", json.loads(response[1]))
print("Inference successful")

# Sample response for sample_chunks[0]

# TitleSummaryJsonResponse :
# {
#   'title': 'Google Cloud Next Conference: Simplifying AI and Machine Learning for Everyone',
#   'summary': 'Google Cloud announced a wide range of innovations and new products in the AI
#               and machine learning space at the recent Google Cloud Next conference. The goal
#               is to make these technologies accessible to everyone by simplifying the process
#               and providing tools for data processing, cybersecurity, and machine learning.
#               Google is also working on advances in AutoML and packaged solutions for certain areas.'
# }