Spaces:
Paused
Paused
add reference code from vllm
Browse files
app.py
CHANGED
@@ -214,65 +214,68 @@ model = load_model(params, model_path)
|
|
214 |
tokenizer = MistralTokenizer.from_model("pixtral")
|
215 |
|
216 |
def preprocess_image(image):
|
|
|
|
|
217 |
image = image.convert('RGB')
|
218 |
image = image.resize((params['vision_encoder']['image_size'], params['vision_encoder']['image_size']))
|
219 |
image_tensor = torch.tensor(np.array(image)).permute(2, 0, 1).unsqueeze(0).float() / 255.0
|
220 |
return image_tensor
|
221 |
|
222 |
-
@spaces.GPU
|
223 |
-
def generate_text(image, prompt):
|
224 |
-
|
225 |
-
|
226 |
-
|
227 |
-
|
228 |
-
|
229 |
-
|
230 |
-
|
231 |
-
|
232 |
-
|
233 |
-
|
234 |
-
|
235 |
-
|
236 |
-
|
|
|
|
|
237 |
)
|
238 |
-
|
239 |
-
|
240 |
-
|
241 |
-
|
242 |
-
|
243 |
-
|
244 |
-
|
245 |
-
|
246 |
-
|
247 |
-
|
248 |
-
|
249 |
-
|
250 |
-
|
251 |
-
|
252 |
-
|
253 |
-
|
254 |
-
|
255 |
-
|
256 |
-
|
257 |
-
@spaces.GPU
|
258 |
def calculate_similarity(image1, image2):
|
259 |
-
|
260 |
-
|
261 |
-
|
262 |
|
263 |
-
|
264 |
-
|
265 |
-
|
266 |
-
|
267 |
-
|
268 |
-
model.cpu()
|
269 |
|
270 |
-
|
271 |
-
similarity = F.cosine_similarity(embedding1, embedding2).item()
|
272 |
|
273 |
-
|
|
|
|
|
274 |
|
275 |
-
with gr.Blocks(
|
276 |
gr.Markdown(title)
|
277 |
gr.Markdown("## Model Details")
|
278 |
gr.Markdown(f"- Model Dimension: {params['dim']}")
|
@@ -287,6 +290,7 @@ with gr.Blocks(theme=gr.themes.Base()) as demo:
|
|
287 |
gr.Markdown("1. The image is processed by a Vision Encoder using 2D ROPE (Rotary Position Embedding).")
|
288 |
gr.Markdown("2. The encoder uses SiLU activation in its feed-forward layers.")
|
289 |
gr.Markdown("3. The encoded image is used for text generation or similarity comparison.")
|
|
|
290 |
gr.Markdown(description)
|
291 |
|
292 |
with gr.Tabs():
|
@@ -295,6 +299,7 @@ with gr.Blocks(theme=gr.themes.Base()) as demo:
|
|
295 |
with gr.Column():
|
296 |
input_image = gr.Image(type="pil", label="Input Image")
|
297 |
input_prompt = gr.Textbox(label="Prompt")
|
|
|
298 |
submit_btn = gr.Button("Generate Text")
|
299 |
|
300 |
with gr.Column():
|
@@ -304,7 +309,7 @@ with gr.Blocks(theme=gr.themes.Base()) as demo:
|
|
304 |
|
305 |
submit_btn.click(
|
306 |
fn=generate_text,
|
307 |
-
inputs=[input_image, input_prompt],
|
308 |
outputs=[output_text, token_count, image_count]
|
309 |
)
|
310 |
|
|
|
214 |
tokenizer = MistralTokenizer.from_model("pixtral")
|
215 |
|
216 |
def preprocess_image(image):
|
217 |
+
if image is None:
|
218 |
+
raise ValueError("No image provided")
|
219 |
image = image.convert('RGB')
|
220 |
image = image.resize((params['vision_encoder']['image_size'], params['vision_encoder']['image_size']))
|
221 |
image_tensor = torch.tensor(np.array(image)).permute(2, 0, 1).unsqueeze(0).float() / 255.0
|
222 |
return image_tensor
|
223 |
|
224 |
+
@spaces.GPU(duration=120)
|
225 |
+
def generate_text(image, prompt, max_tokens):
|
226 |
+
try:
|
227 |
+
image_tensor = preprocess_image(image).cuda()
|
228 |
+
|
229 |
+
tokenized = tokenizer.encode_chat_completion(
|
230 |
+
ChatCompletionRequest(
|
231 |
+
messages=[
|
232 |
+
UserMessage(
|
233 |
+
content=[
|
234 |
+
TextChunk(text=prompt),
|
235 |
+
ImageChunk(image=image),
|
236 |
+
]
|
237 |
+
)
|
238 |
+
],
|
239 |
+
model="pixtral",
|
240 |
+
)
|
241 |
)
|
242 |
+
input_ids = torch.tensor(tokenized.tokens).unsqueeze(0).cuda()
|
243 |
+
|
244 |
+
with torch.no_grad():
|
245 |
+
model.cuda()
|
246 |
+
for _ in range(max_tokens):
|
247 |
+
logits = model(image_tensor, input_ids)
|
248 |
+
next_token_logits = logits[0, -1, :]
|
249 |
+
next_token = torch.argmax(next_token_logits, dim=-1)
|
250 |
+
input_ids = torch.cat([input_ids, next_token.unsqueeze(0).unsqueeze(0)], dim=-1)
|
251 |
+
if next_token.item() == tokenizer.eos_token_id:
|
252 |
+
break
|
253 |
+
model.cpu()
|
254 |
+
|
255 |
+
generated_text = tokenizer.decode(input_ids[0].tolist())
|
256 |
+
return generated_text, len(input_ids[0]), 1 # 1 image processed
|
257 |
+
except Exception as e:
|
258 |
+
return f"Error: {str(e)}", 0, 0
|
259 |
+
|
260 |
+
@spaces.GPU(duration=60)
|
|
|
261 |
def calculate_similarity(image1, image2):
|
262 |
+
try:
|
263 |
+
tensor1 = preprocess_image(image1).cuda()
|
264 |
+
tensor2 = preprocess_image(image2).cuda()
|
265 |
|
266 |
+
with torch.no_grad():
|
267 |
+
model.cuda()
|
268 |
+
embedding1 = model(tensor1).mean(dim=1) # Average over spatial dimensions
|
269 |
+
embedding2 = model(tensor2).mean(dim=1)
|
270 |
+
model.cpu()
|
|
|
271 |
|
272 |
+
similarity = F.cosine_similarity(embedding1, embedding2).item()
|
|
|
273 |
|
274 |
+
return similarity
|
275 |
+
except Exception as e:
|
276 |
+
return f"Error: {str(e)}"
|
277 |
|
278 |
+
with gr.Blocks() as demo:
|
279 |
gr.Markdown(title)
|
280 |
gr.Markdown("## Model Details")
|
281 |
gr.Markdown(f"- Model Dimension: {params['dim']}")
|
|
|
290 |
gr.Markdown("1. The image is processed by a Vision Encoder using 2D ROPE (Rotary Position Embedding).")
|
291 |
gr.Markdown("2. The encoder uses SiLU activation in its feed-forward layers.")
|
292 |
gr.Markdown("3. The encoded image is used for text generation or similarity comparison.")
|
293 |
+
|
294 |
gr.Markdown(description)
|
295 |
|
296 |
with gr.Tabs():
|
|
|
299 |
with gr.Column():
|
300 |
input_image = gr.Image(type="pil", label="Input Image")
|
301 |
input_prompt = gr.Textbox(label="Prompt")
|
302 |
+
max_tokens_slider = gr.Slider(minimum=60, maximum=1600, value=90, step=5, label="Max Tokens")
|
303 |
submit_btn = gr.Button("Generate Text")
|
304 |
|
305 |
with gr.Column():
|
|
|
309 |
|
310 |
submit_btn.click(
|
311 |
fn=generate_text,
|
312 |
+
inputs=[input_image, input_prompt, max_tokens_slider],
|
313 |
outputs=[output_text, token_count, image_count]
|
314 |
)
|
315 |
|