File size: 2,404 Bytes
aecd1c4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
22512a3
aecd1c4
 
93abc8e
aecd1c4
93abc8e
aecd1c4
 
 
 
22512a3
93abc8e
 
8038a90
bed7948
aecd1c4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
22512a3
aecd1c4
 
 
 
 
 
 
 
 
 
 
 
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
import time
from threading import Thread

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

import spaces

model_id = "xtuner/llava-llama-3-8b-v1_1-transformers"

processor = AutoProcessor.from_pretrained(model_id)

model = LlavaForConditionalGeneration.from_pretrained(
    model_id,
    torch_dtype=torch.float16,
    low_cpu_mem_usage=True,
)

model.to("cuda:0")
model.generation_config.eos_token_id = 128009

@spaces.GPU
def bot_streaming(message, history):
    print(message)
    image = None
    if message["files"]:
        if type(message["files"][-1]) == dict:
            image = message["files"][-1]["path"]
        else:
            image = message["files"][-1]
    else:
        for hist in history:
            if type(hist[0]) == tuple:
                image = hist[0][0]
                break

    if image is None:
        image = "ignore.png"
    
    prompt = f"<|start_header_id|>user<|end_header_id|>\n\n<image>\n{message['text']}<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n"
    image = Image.open(image)
    inputs = processor(prompt, image, return_tensors='pt').to(0, torch.float16)

    streamer = TextIteratorStreamer(processor, **{"skip_special_tokens": False, "skip_prompt": True})
    generation_kwargs = dict(inputs, streamer=streamer, max_new_tokens=1024, do_sample=False)

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

    text_prompt = f"<|start_header_id|>user<|end_header_id|>\n\n{message['text']}<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n"

    buffer = ""
    time.sleep(0.5)
    for new_text in streamer:
        if "<|eot_id|>" in new_text:
            new_text = new_text.split("<|eot_id|>")[0]
        buffer += new_text

        generated_text_without_prompt = buffer
        time.sleep(0.06)
        yield generated_text_without_prompt

chatbot=gr.Chatbot(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,
    stop_btn="Stop Generation",
    multimodal=True,
    textbox=chat_input,
    chatbot=chatbot,
    )

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