app.py

import gradio as gr
import spaces
import torch
from torch.cuda.amp import autocast
import subprocess
from huggingface_hub import InferenceClient
import os
import psutil



"""
For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
"""

from accelerate import init_empty_weights, infer_auto_device_map, load_checkpoint_and_dispatch
from accelerate import Accelerator


subprocess.run(
    "pip install psutil",
   
    shell=True,
)

import bitsandbytes as bnb  # Import bitsandbytes for 8-bit quantization



from datetime import datetime


subprocess.run(
    "pip install flash-attn --no-build-isolation",
    env={"FLASH_ATTENTION_SKIP_CUDA_BUILD": "TRUE"},
    shell=True,
)

client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
# pip install 'git+https://github.com/huggingface/transformers.git'



token=os.getenv('token')
print('token = ',token)

from transformers import AutoModelForCausalLM, AutoTokenizer

# model_id = "mistralai/Mistral-7B-v0.3"

model_id = "microsoft/Phi-3-medium-4k-instruct"
# model_id = "Qwen/Qwen2-7B-Instruct"


tokenizer = AutoTokenizer.from_pretrained(
    # model_id
    model_id,
    # use_fast=False
    token= token,
trust_remote_code=True)


accelerator = Accelerator()

model = AutoModelForCausalLM.from_pretrained(model_id, token= token, 
                                                 # torch_dtype= torch.uint8, 
                                             torch_dtype=torch.bfloat16,
                                              # load_in_8bit=True,
                                             # #  # torch_dtype=torch.fl,
                                             attn_implementation="flash_attention_2",
                                             low_cpu_mem_usage=True,
                                             trust_remote_code=True,
                                             device_map='cuda',
                                             # device_map=accelerator.device_map,
                                             
                                            )





# 
model = accelerator.prepare(model)
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline

pipe = pipeline(
    "text-generation",
    model=model,
    tokenizer=tokenizer,
)


# device_map = infer_auto_device_map(model, max_memory={0: "79GB", "cpu":"65GB" })

# Load the model with the inferred device map
# model = load_checkpoint_and_dispatch(model, model_id, device_map=device_map, no_split_module_classes=["GPTJBlock"])
# model.half()

import json

def str_to_json(str_obj):
    json_obj = json.loads(str_obj)
    return json_obj


@spaces.GPU(duration=90)
def respond(
    message,
    history: list[tuple[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
):
    # yield 'retuend'
    # model.to(accelerator.device)

    messages = []
    json_obj = str_to_json(message)
    print(json_obj)
    
    messages= json_obj

    # input_ids = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=True, return_tensors="pt").to(accelerator.device)
    # input_ids2 = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True, return_tensors="pt") #.to('cuda')
    # print(f"Converted input_ids dtype: {input_ids.dtype}")
    # input_str= str(input_ids2)
    # print('input str = ', input_str)

    generation_args = {
    "max_new_tokens": max_tokens,
    "return_full_text": False,
    "temperature": temperature,
    "do_sample": False,
}

    output = pipe(messages, **generation_args)
    print(output[0]['generated_text'])
    gen_text=output[0]['generated_text']

    # with torch.no_grad():
    #     gen_tokens = model.generate(
    # input_ids, 
    # max_new_tokens=max_tokens, 
    # # do_sample=True, 
    # temperature=temperature,
    # )

    # gen_text = tokenizer.decode(gen_tokens[0])
    # print(gen_text)
    # gen_text= gen_text.replace(input_str,'')
    # gen_text= gen_text.replace('<|im_end|>','')
    
    yield gen_text
   
  
#     messages = [
#     # {"role": "user", "content": "What is your favourite condiment?"},
#     # {"role": "assistant", "content": "Well, I'm quite partial to a good squeeze of fresh lemon juice. It adds just the right amount of zesty flavour to whatever I'm cooking up in the kitchen!"},
#     # {"role": "user", "content": "Do you have mayonnaise recipes?"}
# ]

    # inputs = tokenizer.apply_chat_template(messages, return_tensors="pt").to("cuda")

    # outputs = model.generate(inputs, max_new_tokens=2000)
    # gen_text=tokenizer.decode(outputs[0], skip_special_tokens=True)
   
    # print(gen_text)
    # yield gen_text
    # for val in history:
    #     if val[0]:
    #         messages.append({"role": "user", "content": val[0]})
    #     if val[1]:
    #         messages.append({"role": "assistant", "content": val[1]})

    # messages.append({"role": "user", "content": message})

    # response = ""

    # for message in client.chat_completion(
    #     messages,
    #     max_tokens=max_tokens,
    #     stream=True,
    #     temperature=temperature,
    #     top_p=top_p,
    # ):
    #     token = message.choices[0].delta.content

    #     response += token
    #     yield response

"""
For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
"""
demo = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
        gr.Slider(
            minimum=0.1,
            maximum=1.0,
            value=0.95,
            step=0.05,
            label="Top-p (nucleus sampling)",
        ),
    ],
)


if __name__ == "__main__":
    demo.launch()