app.py

import gradio as gr
from transformers import TextIteratorStreamer
from threading import Thread
from transformers import StoppingCriteria, StoppingCriteriaList
import torch
import spaces
import os
import subprocess

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

theme = gr.themes.Base(
    font=[gr.themes.GoogleFont('Libre Franklin'), gr.themes.GoogleFont('Public Sans'), 'system-ui', 'sans-serif'],
)

model_name = "microsoft/Phi-3-medium-4k-instruct"
from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained(model_name, device_map='cuda', torch_dtype=torch.float16, _attn_implementation="flash_attention_2", trust_remote_code=True)
tokenizer = AutoTokenizer.from_pretrained(model_name)

class StopOnTokens(StoppingCriteria):
    def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor, **kwargs) -> bool:
        stop_ids = [29, 0]
        for stop_id in stop_ids:
            if input_ids[0][-1] == stop_id:
                return True
        return False

@spaces.GPU(queue=False)
def predict1(message, history, temperature1, max_tokens1, repetition_penalty1, top_p1):
    history_transformer_format = history + [[message, ""]]
    stop = StopOnTokens()
    messages = "".join(["".join(["\n<|end|>\n<|user|>\n"+item[0], "\n<|end|>\n<|assistant|>\n"+item[1]]) for item in history_transformer_format])
    model_inputs = tokenizer([messages], return_tensors="pt").to("cuda")
    streamer = TextIteratorStreamer(tokenizer, timeout=10., skip_prompt=True, skip_special_tokens=True)
    generate_kwargs = dict(
        model_inputs,
        streamer=streamer,
        max_new_tokens=max_tokens1,
        do_sample=True,
        top_p=top_p1,
        repetition_penalty=repetition_penalty1,
        temperature=temperature1,
        stopping_criteria=StoppingCriteriaList([stop])
    )
    t = Thread(target=model.generate, kwargs=generate_kwargs)
    t.start()
    partial_message = ""
    for new_token in streamer:
        if new_token != '<':
            partial_message += new_token
        yield partial_message



model_name = "microsoft/Phi-3-medium-128k-instruct"
from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained(model_name, device_map='cuda', torch_dtype=torch.float16, _attn_implementation="flash_attention_2", trust_remote_code=True)
tokenizer = AutoTokenizer.from_pretrained(model_name)

class StopOnTokens(StoppingCriteria):
    def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor, **kwargs) -> bool:
        stop_ids = [29, 0]
        for stop_id in stop_ids:
            if input_ids[0][-1] == stop_id:
                return True
        return False

@spaces.GPU(queue=False)
def predict(message, history, temperature, max_tokens, repetition_penalty, top_p):
    history_transformer_format = history + [[message, ""]]
    stop = StopOnTokens()
    messages = "".join(["".join(["\n<|end|>\n<|user|>\n"+item[0], "\n<|end|>\n<|assistant|>\n"+item[1]]) for item in history_transformer_format])
    model_inputs = tokenizer([messages], return_tensors="pt").to("cuda")
    streamer = TextIteratorStreamer(tokenizer, timeout=10., skip_prompt=True, skip_special_tokens=True)
    generate_kwargs = dict(
        model_inputs,
        streamer=streamer,
        max_new_tokens=max_tokens,
        do_sample=True,
        top_p=top_p,
        repetition_penalty=repetition_penalty,
        temperature=temperature,
        stopping_criteria=StoppingCriteriaList([stop])
    )
    t = Thread(target=model.generate, kwargs=generate_kwargs)
    t.start()
    partial_message = ""
    for new_token in streamer:
        if new_token != '<':
            partial_message += new_token
        yield partial_message

max_tokens1 = gr.Slider(
    minimum=512,
    maximum=4096,
    value=4000,
    step=32,
    interactive=True,
    label="Maximum number of new tokens to generate",
)
repetition_penalty1 = gr.Slider(
    minimum=0.01,
    maximum=5.0,
    value=1,
    step=0.01,
    interactive=True,
    label="Repetition penalty",
)
temperature1 = gr.Slider(
    minimum=0.0,
    maximum=1.0,
    value=0.7,
    step=0.05,
    visible=True,
    interactive=True,
    label="Temperature",
)
top_p1 = gr.Slider(
    minimum=0.01,
    maximum=0.99,
    value=0.9,
    step=0.01,
    visible=True,
    interactive=True,
    label="Top P",
)

chatbot1 = gr.Chatbot(
    label="Phi3-medium-4k",
    show_copy_button=True, 
    likeable=True, 
    layout="panel"
)

output=gr.Textbox(label="Prompt")

with gr.Blocks() as min:  
    gr.ChatInterface(
    fn=predict1,
    chatbot=chatbot1,
    additional_inputs=[
        temperature1,
        max_tokens1,
        repetition_penalty1,
        top_p1,
        ],   
    )

max_tokens = gr.Slider(
    minimum=64000,
    maximum=128000,
    value=100000,
    step=1000,
    interactive=True,
    label="Maximum number of new tokens to generate",
)
repetition_penalty = gr.Slider(
    minimum=0.01,
    maximum=5.0,
    value=1,
    step=0.01,
    interactive=True,
    label="Repetition penalty",
)
temperature = gr.Slider(
    minimum=0.0,
    maximum=1.0,
    value=0.7,
    step=0.05,
    visible=True,
    interactive=True,
    label="Temperature",
)
top_p = gr.Slider(
    minimum=0.01,
    maximum=0.99,
    value=0.9,
    step=0.01,
    visible=True,
    interactive=True,
    label="Top P",
)

chatbot = gr.Chatbot(
    label="Phi3-medium-128k",
    show_copy_button=True, 
    likeable=True, 
    layout="panel"
)

output=gr.Textbox(label="Prompt")
    
with gr.Blocks() as max:
    gr.ChatInterface(
    fn=predict,
    chatbot=chatbot,
    additional_inputs=[
        temperature,
        max_tokens,
        repetition_penalty,
        top_p,
        ],   
    )

with gr.Blocks(title="Phi 3 Medium DEMO", theme=theme) as demo:
    gr.Markdown("# Phi3 Medium all in one")
    gr.TabbedInterface([max, min], ['Phi3 medium 128k','Phi3 medium 4k'])
        
demo.launch(share=True)