Hugging Face's logo

Gokulram2710
/
Phi3_vision

Model card Files Community
Phi3_vision
File size: 5,008 Bytes 
28b96b3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
 
import time
from threading import Thread

import gradio as gr
import spaces
import torch
from PIL import Image
from transformers import AutoProcessor, AutoModelForCausalLM
from transformers import TextIteratorStreamer

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

# thanks to https://huggingface.co./ysharma
PLACEHOLDER = """

<div style="padding: 30px; text-align: center; display: flex; flex-direction: column; align-items: center;">

   <img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/models/microsoft/Phi-3-vision-128k-instruct.png" style="width: 80%; max-width: 550px; height: auto; opacity: 0.55;  "> 

   <h1 style="font-size: 28px; margin-bottom: 2px; opacity: 0.55;">Microsoft's Phi3-Vision-128k-Context</h1>

   <p style="font-size: 18px; margin-bottom: 2px; opacity: 0.65;">Phi-3-Vision is a 4.2B parameter multimodal model that brings together language and vision capabilities.</p>

</div>

"""

user_prompt = '<|user|>\n'
assistant_prompt = '<|assistant|>\n'
prompt_suffix = "<|end|>\n"



model_id = "microsoft/Phi-3-vision-128k-instruct"

processor = AutoProcessor.from_pretrained(model_id, trust_remote_code=True)

model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype="auto",
    trust_remote_code=True,
)

model.to("cuda:0")


@spaces.GPU
def bot_streaming(message, history):
    print(f'message is - {message}')
    print(f'history is - {history}')
    if message["files"]:
        # message["files"][-1] is a Dict or just a string
        if type(message["files"][-1]) == dict:
            image = message["files"][-1]["path"]
        else:
            image = message["files"][-1]
    else:
        # if there's no image uploaded for this turn, look for images in the past turns
        # kept inside tuples, take the last one
        for hist in history:
            if type(hist[0]) == tuple:
                image = hist[0][0]
    try:
        if image is None:
            # Handle the case where image is None
            raise gr.Error("You need to upload an image for Phi3-Vision to work. Close the error and try again with an Image.")
    except NameError:
        # Handle the case where 'image' is not defined at all
        raise gr.Error("You need to upload an image for Phi3-Vision to work. Close the error and try again with an Image.")

    conversation = []
    flag=False
    for user, assistant in history:
        if assistant is None:
            #pass
            flag=True
            conversation.extend([{"role": "user", "content":""}])
            continue
        if flag==True:
            conversation[0]['content'] = f"<|image_1|>\n{user}"   
            conversation.extend([{"role": "assistant", "content": assistant}])
            flag=False
            continue
        conversation.extend([{"role": "user", "content": user}, {"role": "assistant", "content": assistant}])

    if len(history) == 0:
        conversation.append({"role": "user", "content": f"<|image_1|>\n{message['text']}"})
    else:
        conversation.append({"role": "user", "content": message['text']})
    print(f"prompt is -\n{conversation}")
    prompt = processor.tokenizer.apply_chat_template(conversation, tokenize=False, add_generation_prompt=True)
    image = Image.open(image)
    inputs = processor(prompt, image, return_tensors="pt").to("cuda:0") 

    streamer = TextIteratorStreamer(processor, **{"skip_special_tokens": True, "skip_prompt": True, 'clean_up_tokenization_spaces':False,}) 
    generation_kwargs = dict(inputs, streamer=streamer, max_new_tokens=1024, do_sample=False, temperature=0.0, eos_token_id=processor.tokenizer.eos_token_id,)

    thread = Thread(target=model.generate, kwargs=generation_kwargs)
    thread.start()

    buffer = ""
    for new_text in streamer:
        buffer += new_text
        yield buffer


chatbot = gr.Chatbot(placeholder=PLACEHOLDER, scale=1)
chat_input = gr.MultimodalTextbox(interactive=True, file_types=["image"], placeholder="Enter message or upload file...",
                                  show_label=False)
with gr.Blocks(fill_height=True, ) as demo:
    gr.ChatInterface(
        fn=bot_streaming,
        title="Phi-3 Vision 128k Instruct",
        examples=[{"text": "What is on the flower?", "files": ["./bee.jpg"]},
                  {"text": "How to make this pastry?", "files": ["./baklava.png"]}],
        description="Try [microsoft/Phi-3-vision-128k-instruct](https://huggingface.co./microsoft/Phi-3-vision-128k-instruct). Upload an image and start chatting about it, or simply try one of the examples below. If you don't upload an image, you will receive an error.",
        stop_btn="Stop Generation",
        multimodal=True,
        textbox=chat_input,
        chatbot=chatbot,
    )

demo.queue(api_open=False)
demo.launch(show_api=False, share=False)