Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -1,22 +1,22 @@
|
|
1 |
-
import
|
|
|
2 |
from llava_llama3.serve.cli import chat_llava
|
3 |
from llava_llama3.model.builder import load_pretrained_model
|
4 |
-
|
5 |
import torch
|
|
|
|
|
6 |
import spaces
|
7 |
|
8 |
# Model configuration
|
9 |
-
|
10 |
-
device = "cuda"
|
11 |
-
conv_mode = "llama_3"
|
12 |
-
temperature = 0
|
13 |
-
max_new_tokens = 512
|
14 |
load_8bit = False
|
15 |
load_4bit = False
|
16 |
|
17 |
# Load the pretrained model
|
18 |
tokenizer, llava_model, image_processor, context_len = load_pretrained_model(
|
19 |
-
|
20 |
None,
|
21 |
'llava_llama3',
|
22 |
load_8bit,
|
@@ -24,38 +24,67 @@ tokenizer, llava_model, image_processor, context_len = load_pretrained_model(
|
|
24 |
device=device
|
25 |
)
|
26 |
|
27 |
-
|
28 |
@spaces.GPU
|
29 |
-
def bot_streaming(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
30 |
output = chat_llava(
|
31 |
args=None,
|
32 |
image_file=image,
|
33 |
-
text=
|
34 |
tokenizer=tokenizer,
|
35 |
model=llava_model,
|
36 |
image_processor=image_processor,
|
37 |
context_len=context_len
|
38 |
)
|
39 |
-
|
40 |
-
|
41 |
-
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
|
49 |
-
|
50 |
-
|
51 |
-
|
52 |
-
|
53 |
-
|
54 |
-
|
55 |
-
|
56 |
-
|
|
|
|
|
|
|
57 |
)
|
58 |
|
59 |
-
# Launch the Gradio app
|
60 |
demo.queue(api_open=False)
|
61 |
-
demo.launch(show_api=False, share=False)
|
|
|
1 |
+
import time
|
2 |
+
from threading import Thread
|
3 |
from llava_llama3.serve.cli import chat_llava
|
4 |
from llava_llama3.model.builder import load_pretrained_model
|
5 |
+
import gradio as gr
|
6 |
import torch
|
7 |
+
from PIL import Image
|
8 |
+
|
9 |
import spaces
|
10 |
|
11 |
# Model configuration
|
12 |
+
model_id = "TheFinAI/FinLLaVA"
|
13 |
+
device = "cuda:0"
|
|
|
|
|
|
|
14 |
load_8bit = False
|
15 |
load_4bit = False
|
16 |
|
17 |
# Load the pretrained model
|
18 |
tokenizer, llava_model, image_processor, context_len = load_pretrained_model(
|
19 |
+
model_id,
|
20 |
None,
|
21 |
'llava_llama3',
|
22 |
load_8bit,
|
|
|
24 |
device=device
|
25 |
)
|
26 |
|
27 |
+
|
28 |
@spaces.GPU
|
29 |
+
def bot_streaming(message, history):
|
30 |
+
print(message)
|
31 |
+
image = None
|
32 |
+
|
33 |
+
# Check if there's an image in the current message
|
34 |
+
if message["files"]:
|
35 |
+
# message["files"][-1] could be a dictionary or a string
|
36 |
+
if isinstance(message["files"][-1], dict):
|
37 |
+
image = message["files"][-1]["path"]
|
38 |
+
else:
|
39 |
+
image = message["files"][-1]
|
40 |
+
else:
|
41 |
+
# If no image in the current message, look in the history for the last image
|
42 |
+
for hist in history:
|
43 |
+
if isinstance(hist[0], tuple):
|
44 |
+
image = hist[0][0]
|
45 |
+
|
46 |
+
# Error handling if no image is found
|
47 |
+
if image is None:
|
48 |
+
raise gr.Error("You need to upload an image for LLaVA to work.")
|
49 |
+
|
50 |
+
# Load the image
|
51 |
+
image = Image.open(image)
|
52 |
+
|
53 |
+
# Generate the prompt for the model
|
54 |
+
prompt = message['text']
|
55 |
+
|
56 |
+
# Call the chat_llava function to generate the output
|
57 |
output = chat_llava(
|
58 |
args=None,
|
59 |
image_file=image,
|
60 |
+
text=prompt,
|
61 |
tokenizer=tokenizer,
|
62 |
model=llava_model,
|
63 |
image_processor=image_processor,
|
64 |
context_len=context_len
|
65 |
)
|
66 |
+
|
67 |
+
# Stream the output
|
68 |
+
buffer = ""
|
69 |
+
for new_text in output:
|
70 |
+
buffer += new_text
|
71 |
+
yield buffer
|
72 |
+
|
73 |
+
|
74 |
+
chatbot=gr.Chatbot(scale=1)
|
75 |
+
chat_input = gr.MultimodalTextbox(interactive=True, file_types=["image"], placeholder="Enter message or upload file...", show_label=False)
|
76 |
+
with gr.Blocks(fill_height=True, ) as demo:
|
77 |
+
gr.ChatInterface(
|
78 |
+
fn=bot_streaming,
|
79 |
+
title="LLaVA Llama-3-8B",
|
80 |
+
examples=[{"text": "What is on the flower?", "files": ["./bee.jpg"]},
|
81 |
+
{"text": "How to make this pastry?", "files": ["./baklava.png"]}],
|
82 |
+
|
83 |
+
stop_btn="Stop Generation",
|
84 |
+
multimodal=True,
|
85 |
+
textbox=chat_input,
|
86 |
+
chatbot=chatbot,
|
87 |
)
|
88 |
|
|
|
89 |
demo.queue(api_open=False)
|
90 |
+
demo.launch(show_api=False, share=False)
|