Update app.py
Browse files
app.py
CHANGED
|
@@ -1,106 +1,12 @@
|
|
| 1 |
import gradio as gr
|
| 2 |
-
import torch
|
| 3 |
-
from threading import Thread
|
| 4 |
-
from transformers import (
|
| 5 |
-
AutoProcessor,
|
| 6 |
-
AutoModelForImageTextToText,
|
| 7 |
-
TextIteratorStreamer,
|
| 8 |
-
)
|
| 9 |
-
|
| 10 |
-
device = "cuda" if torch.cuda.is_available() else "cpu"
|
| 11 |
|
| 12 |
-
|
| 13 |
-
"
|
| 14 |
-
)
|
| 15 |
|
| 16 |
-
|
| 17 |
-
|
| 18 |
-
|
|
|
|
| 19 |
)
|
| 20 |
-
model.to(device)
|
| 21 |
-
|
| 22 |
-
|
| 23 |
-
def infer(text, files, history, max_tokens):
|
| 24 |
-
user_content = []
|
| 25 |
-
|
| 26 |
-
if text.strip():
|
| 27 |
-
user_content.append({"type": "text", "text": text})
|
| 28 |
-
|
| 29 |
-
if files:
|
| 30 |
-
for f in files:
|
| 31 |
-
name = f.name
|
| 32 |
-
if name.lower().endswith((".png", ".jpg", ".jpeg")):
|
| 33 |
-
user_content.append({"type": "image", "path": name})
|
| 34 |
-
elif name.lower().endswith(".mp4"):
|
| 35 |
-
user_content.append({"type": "video", "path": name})
|
| 36 |
-
|
| 37 |
-
messages = [{"role": "user", "content": user_content}]
|
| 38 |
-
|
| 39 |
-
inputs = processor.apply_chat_template(
|
| 40 |
-
messages,
|
| 41 |
-
add_generation_prompt=True,
|
| 42 |
-
tokenize=True,
|
| 43 |
-
return_tensors="pt",
|
| 44 |
-
).to(device)
|
| 45 |
-
|
| 46 |
-
streamer = TextIteratorStreamer(
|
| 47 |
-
processor, skip_prompt=True, skip_special_tokens=True
|
| 48 |
-
)
|
| 49 |
-
|
| 50 |
-
thread = Thread(
|
| 51 |
-
target=model.generate,
|
| 52 |
-
kwargs=dict(
|
| 53 |
-
inputs,
|
| 54 |
-
streamer=streamer,
|
| 55 |
-
max_new_tokens=max_tokens,
|
| 56 |
-
),
|
| 57 |
-
)
|
| 58 |
-
thread.start()
|
| 59 |
-
|
| 60 |
-
output = ""
|
| 61 |
-
for token in streamer:
|
| 62 |
-
output += token
|
| 63 |
-
yield output
|
| 64 |
-
|
| 65 |
-
|
| 66 |
-
with gr.Blocks() as demo:
|
| 67 |
-
gr.Markdown("## 🧠 SmolVLM2 – Stable Space UI")
|
| 68 |
-
|
| 69 |
-
chatbot = gr.Chatbot(height=450)
|
| 70 |
-
|
| 71 |
-
text = gr.Textbox(
|
| 72 |
-
label="Your question",
|
| 73 |
-
placeholder="Ask something about the image or video",
|
| 74 |
-
)
|
| 75 |
-
|
| 76 |
-
files = gr.Files(
|
| 77 |
-
label="Upload image or video",
|
| 78 |
-
file_types=["image", ".mp4"],
|
| 79 |
-
)
|
| 80 |
-
|
| 81 |
-
max_tokens = gr.Slider(50, 400, value=200, step=50)
|
| 82 |
-
|
| 83 |
-
def user_submit(t, f, history):
|
| 84 |
-
history = history + [[t, None]]
|
| 85 |
-
return "", [], history
|
| 86 |
-
|
| 87 |
-
def bot_reply(history, max_tokens):
|
| 88 |
-
last_text = history[-1][0]
|
| 89 |
-
for chunk in infer(last_text, files.value, history[:-1], max_tokens):
|
| 90 |
-
history[-1][1] = chunk
|
| 91 |
-
yield history
|
| 92 |
-
|
| 93 |
-
submit = gr.Button("Send")
|
| 94 |
-
|
| 95 |
-
submit.click(
|
| 96 |
-
user_submit,
|
| 97 |
-
[text, files, chatbot],
|
| 98 |
-
[text, files, chatbot],
|
| 99 |
-
queue=False,
|
| 100 |
-
).then(
|
| 101 |
-
bot_reply,
|
| 102 |
-
[chatbot, max_tokens],
|
| 103 |
-
chatbot,
|
| 104 |
-
)
|
| 105 |
|
| 106 |
-
demo.launch(
|
|
|
|
| 1 |
import gradio as gr
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
|
| 3 |
+
def hello(x):
|
| 4 |
+
return f"Hello {x}"
|
|
|
|
| 5 |
|
| 6 |
+
demo = gr.Interface(
|
| 7 |
+
fn=hello,
|
| 8 |
+
inputs=gr.Textbox(),
|
| 9 |
+
outputs=gr.Textbox(),
|
| 10 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 11 |
|
| 12 |
+
demo.launch()
|