Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -20,7 +20,6 @@ from transformers import (
|
|
20 |
TextIteratorStreamer,
|
21 |
Qwen2VLForConditionalGeneration,
|
22 |
AutoProcessor,
|
23 |
-
AutoModelForImageTextToText, # <-- New import for aya-vision
|
24 |
)
|
25 |
from transformers.image_utils import load_image
|
26 |
from diffusers import StableDiffusionXLPipeline, EulerAncestralDiscreteScheduler
|
@@ -52,16 +51,6 @@ model_m = Qwen2VLForConditionalGeneration.from_pretrained(
|
|
52 |
torch_dtype=torch.float16
|
53 |
).to("cuda").eval()
|
54 |
|
55 |
-
# --- New feature: aya-vision ---
|
56 |
-
AYA_MODEL_ID = "CohereForAI/aya-vision-8b"
|
57 |
-
aya_processor = AutoProcessor.from_pretrained(AYA_MODEL_ID, trust_remote_code=True)
|
58 |
-
aya_model = AutoModelForImageTextToText.from_pretrained(
|
59 |
-
AYA_MODEL_ID,
|
60 |
-
trust_remote_code=True,
|
61 |
-
torch_dtype=torch.float16
|
62 |
-
).to("cuda").eval()
|
63 |
-
# --------------------------------
|
64 |
-
|
65 |
async def text_to_speech(text: str, voice: str, output_file="output.mp3"):
|
66 |
communicate = edge_tts.Communicate(text, voice)
|
67 |
await communicate.save(output_file)
|
@@ -199,38 +188,6 @@ def generate(
|
|
199 |
files = input_dict.get("files", [])
|
200 |
|
201 |
lower_text = text.lower().strip()
|
202 |
-
|
203 |
-
# --- New branch for @aya-vision feature ---
|
204 |
-
if lower_text.startswith("@aya-vision"):
|
205 |
-
prompt_clean = re.sub(r"@aya-vision", "", text, flags=re.IGNORECASE).strip().strip('"')
|
206 |
-
if not files:
|
207 |
-
yield "Please provide an image for @aya-vision command."
|
208 |
-
return
|
209 |
-
image = load_image(files[0])
|
210 |
-
messages = [{
|
211 |
-
"role": "user",
|
212 |
-
"content": [
|
213 |
-
{"type": "image", "image": image},
|
214 |
-
{"type": "text", "text": prompt_clean},
|
215 |
-
]
|
216 |
-
}]
|
217 |
-
prompt_aya = aya_processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
|
218 |
-
inputs = aya_processor(text=[prompt_aya], images=[image], return_tensors="pt", padding=True).to("cuda")
|
219 |
-
streamer = TextIteratorStreamer(aya_processor, skip_prompt=True, skip_special_tokens=True)
|
220 |
-
generation_kwargs = {**inputs, "streamer": streamer, "max_new_tokens": max_new_tokens}
|
221 |
-
thread = Thread(target=aya_model.generate, kwargs=generation_kwargs)
|
222 |
-
thread.start()
|
223 |
-
|
224 |
-
buffer = ""
|
225 |
-
yield "💭 Processing @aya-vision..."
|
226 |
-
for new_text in streamer:
|
227 |
-
buffer += new_text
|
228 |
-
buffer = buffer.replace("<|im_end|>", "")
|
229 |
-
time.sleep(0.01)
|
230 |
-
yield buffer
|
231 |
-
return
|
232 |
-
# ------------------------------------------------
|
233 |
-
|
234 |
# Check if the prompt is an image generation command using model flags.
|
235 |
if (lower_text.startswith("@lightningv5") or
|
236 |
lower_text.startswith("@lightningv4") or
|
@@ -382,22 +339,19 @@ demo = gr.ChatInterface(
|
|
382 |
gr.Slider(label="Repetition penalty", minimum=1.0, maximum=2.0, step=0.05, value=1.2),
|
383 |
],
|
384 |
examples=[
|
385 |
-
[{"text": "@aya-vision Extract JSON from the image", "files": ["examples/document.jpg"]}],
|
386 |
-
[{"text": "@aya-vision Summarize the letter", "files": ["examples/1.png"]}],
|
387 |
["Python Program for Array Rotation"],
|
388 |
["@tts1 Who is Nikola Tesla, and why did he die?"],
|
389 |
['@lightningv5 Chocolate dripping from a donut against a yellow background, in the style of brocore, hyper-realistic'],
|
390 |
['@lightningv4 A serene landscape with mountains'],
|
391 |
['@turbov3 Abstract art, colorful and vibrant'],
|
392 |
["@tts2 What causes rainbows to form?"],
|
393 |
-
[" Describe the content of this image"],
|
394 |
],
|
395 |
cache_examples=False,
|
396 |
type="messages",
|
397 |
description=DESCRIPTION,
|
398 |
css=css,
|
399 |
fill_height=True,
|
400 |
-
textbox=gr.MultimodalTextbox(label="Query Input", file_types=["image"], file_count="multiple", placeholder="
|
401 |
stop_btn="Stop Generation",
|
402 |
multimodal=True,
|
403 |
|
|
|
20 |
TextIteratorStreamer,
|
21 |
Qwen2VLForConditionalGeneration,
|
22 |
AutoProcessor,
|
|
|
23 |
)
|
24 |
from transformers.image_utils import load_image
|
25 |
from diffusers import StableDiffusionXLPipeline, EulerAncestralDiscreteScheduler
|
|
|
51 |
torch_dtype=torch.float16
|
52 |
).to("cuda").eval()
|
53 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
54 |
async def text_to_speech(text: str, voice: str, output_file="output.mp3"):
|
55 |
communicate = edge_tts.Communicate(text, voice)
|
56 |
await communicate.save(output_file)
|
|
|
188 |
files = input_dict.get("files", [])
|
189 |
|
190 |
lower_text = text.lower().strip()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
191 |
# Check if the prompt is an image generation command using model flags.
|
192 |
if (lower_text.startswith("@lightningv5") or
|
193 |
lower_text.startswith("@lightningv4") or
|
|
|
339 |
gr.Slider(label="Repetition penalty", minimum=1.0, maximum=2.0, step=0.05, value=1.2),
|
340 |
],
|
341 |
examples=[
|
|
|
|
|
342 |
["Python Program for Array Rotation"],
|
343 |
["@tts1 Who is Nikola Tesla, and why did he die?"],
|
344 |
['@lightningv5 Chocolate dripping from a donut against a yellow background, in the style of brocore, hyper-realistic'],
|
345 |
['@lightningv4 A serene landscape with mountains'],
|
346 |
['@turbov3 Abstract art, colorful and vibrant'],
|
347 |
["@tts2 What causes rainbows to form?"],
|
|
|
348 |
],
|
349 |
cache_examples=False,
|
350 |
type="messages",
|
351 |
description=DESCRIPTION,
|
352 |
css=css,
|
353 |
fill_height=True,
|
354 |
+
textbox=gr.MultimodalTextbox(label="Query Input", file_types=["image"], file_count="multiple", placeholder="use the tags @lightningv5 @lightningv4 @turbov3 for image gen !"),
|
355 |
stop_btn="Stop Generation",
|
356 |
multimodal=True,
|
357 |
|