app.py

from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
import gradio as gr
import torch
from peft import PeftConfig, PeftModel

# Loading PEFT model
PEFT_MODEL = "TurtleLiu/mistral7b_psychology_bot"

config = PeftConfig.from_pretrained(PEFT_MODEL)

bnb_config = BitsAndBytesConfig(
    load_in_4bit= True,
    bnb_4bit_quant_type= "nf4",
    bnb_4bit_compute_dtype= torch.bfloat16,
    bnb_4bit_use_double_quant= False,
)

peft_base_model = AutoModelForCausalLM.from_pretrained(
    config.base_model_name_or_path,
    return_dict=True,
    quantization_config=bnb_config,
    device_map="auto",
    trust_remote_code=True,
)

model = PeftModel.from_pretrained(peft_base_model, PEFT_MODEL)
model = model.merge_and_unload()

# Load tokenizer
tokenizer = AutoTokenizer.from_pretrained(config.base_model_name_or_path, trust_remote_code=True)
tokenizer.pad_token = tokenizer.eos_token
tokenizer.padding_side = "right"

# Generate response

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

pipe = pipeline(task="text-generation", model=model, tokenizer=tokenizer, max_length=200, do_sample=True,
    max_new_tokens=1024, 
    temperature=0.9, 
    top_k=50, 
    top_p=0.95,
    num_return_sequences=1)

def generate_response(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]"
    result = pipe(f"{prompt}")[0]['generated_text']
    return result

'''
def generate_response(prompt, history, temperature=0.9, max_new_tokens=1024, top_p=0.95, repetition_penalty=1.0, **kwargs,):
    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,
    )
    runtimeFlag = "cuda:0"
    formatted_prompt = format_prompt(f"{prompt}", history)
    inputs = tokenizer([formatted_prompt], return_tensors="pt").to(runtimeFlag)
    generation_config = GenerationConfig(
        temperature=temperature,
        top_p=top_p,
        max_new_tokens=max_new_tokens,
        repetition_penalty=repetition_penalty,
        do_sample=True,
        **kwargs,
    )
    generation_output = model.generate(
            **inputs,
            generation_config=generation_config,
            return_dict_in_generate=True,
            output_scores=True,
            max_new_tokens=max_new_tokens,
        )
'''

# UI design
examples=[
    ["Patient is feeling stressed due to work and has trouble sleeping.", None, None, None, None, None],
    ["Client is dealing with relationship issues and is seeking advice on communication strategies.", None, None, None, None, None],
    ["Individual has recently experienced a loss and is having difficulty coping with grief.", None, None, None, None, None],
]

gr.ChatInterface(
    fn=generate_response,
    chatbot=gr.Chatbot(show_label=False, show_share_button=False, show_copy_button=True, likeable=True, layout="panel"),
    title="Psychological Assistant: Expert in Assessment and Strategic Planning",
    description="Enter counseling notes to generate an assessment and plan.",
    examples=examples,
    concurrency_limit=20,
).launch(show_api=False, debug=True)
    

'''
from huggingface_hub import InferenceClient
import gradio as gr

client = InferenceClient(
    "TurtleLiu/mistral7b_psychology_bot"
)


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 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] As a psychology counselor assistant, provide an assessment and plan for the following counseling notes. Please present a summary, don't make it so long. Present in lines.: {message} [/INST]"
    return prompt

    
def generate(
    prompt, history, temperature=0.9, max_new_tokens=1024, 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(f"{prompt}", history)
    stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
    output = ""

    for response in stream:
        output += response.token.text
        yield output
    return output


examples=[
    ["Patient is feeling stressed due to work and has trouble sleeping.", None, None, None, None, None],
    ["Client is dealing with relationship issues and is seeking advice on communication strategies.", None, None, None, None, None],
    ["Individual has recently experienced a loss and is having difficulty coping with grief.", None, None, None, None, None],
]

gr.ChatInterface(
    fn=generate,
    chatbot=gr.Chatbot(show_label=False, show_share_button=False, show_copy_button=True, likeable=True, layout="panel"),
    title="Psychological Assistant: Expert in Assessment and Strategic Planning",
    description="Enter counseling notes to generate an assessment and plan.",
    examples=examples,
    concurrency_limit=20,
).launch(show_api=False, debug=True)
'''