cnzzx commited on
Commit
c640227
1 Parent(s): 686a863
Files changed (2) hide show
  1. app.py +2 -2
  2. models/vsa_model.py +22 -15
app.py CHANGED
@@ -100,8 +100,8 @@ with gr.Blocks() as app:
100
 
101
 
102
  vsa = VisionSearchAssistant(
103
- ground_device = "cuda:0",
104
- vlm_device="cuda:0",
105
  vlm_load_4bit=True,
106
  )
107
  # Launch the app
 
100
 
101
 
102
  vsa = VisionSearchAssistant(
103
+ ground_device = "cuda",
104
+ vlm_device="cuda",
105
  vlm_load_4bit=True,
106
  )
107
  # Launch the app
models/vsa_model.py CHANGED
@@ -278,7 +278,6 @@ class VisionSearchAssistant:
278
  the $\mathcal{F}_{vlm}(cdot)$ in the paper.
279
 
280
  """
281
- @spaces.GPU
282
  def __init__(
283
  self,
284
  search_model: str = "internlm/internlm2_5-7b-chat",
@@ -289,19 +288,13 @@ class VisionSearchAssistant:
289
  vlm_load_4bit: bool = True,
290
  vlm_load_8bit: bool = False,
291
  ):
292
- self.searcher = WebSearcher(
293
- model_path = search_model
294
- )
295
- self.grounder = VisualGrounder(
296
- model_path = ground_model,
297
- device = ground_device,
298
- )
299
- self.vlm = VLM(
300
- model_path = vlm_model,
301
- device = vlm_device,
302
- load_4bit = vlm_load_4bit,
303
- load_8bit = vlm_load_8bit
304
- )
305
  self.use_correlate = True
306
 
307
  @spaces.GPU
@@ -310,7 +303,21 @@ class VisionSearchAssistant:
310
  image: Union[str, Image.Image, np.ndarray],
311
  text: str,
312
  ground_classes: Union[List[str], None] = None
313
- ):
 
 
 
 
 
 
 
 
 
 
 
 
 
 
314
  # Create and clear the temporary directory.
315
  if not os.access('temp', os.F_OK):
316
  os.makedirs('temp')
 
278
  the $\mathcal{F}_{vlm}(cdot)$ in the paper.
279
 
280
  """
 
281
  def __init__(
282
  self,
283
  search_model: str = "internlm/internlm2_5-7b-chat",
 
288
  vlm_load_4bit: bool = True,
289
  vlm_load_8bit: bool = False,
290
  ):
291
+ self.search_model = search_model
292
+ self.ground_model = ground_model
293
+ self.ground_device = ground_device
294
+ self.vlm_model = vlm_model
295
+ self.vlm_device = vlm_device
296
+ self.vlm_load_4bit = vlm_load_4bit
297
+ self.vlm_load_8bit = vlm_load_8bit
 
 
 
 
 
 
298
  self.use_correlate = True
299
 
300
  @spaces.GPU
 
303
  image: Union[str, Image.Image, np.ndarray],
304
  text: str,
305
  ground_classes: Union[List[str], None] = None
306
+ ):
307
+ self.searcher = WebSearcher(
308
+ model_path = self.search_model
309
+ )
310
+ self.grounder = VisualGrounder(
311
+ model_path = self.ground_model,
312
+ device = self.ground_device,
313
+ )
314
+ self.vlm = VLM(
315
+ model_path = self.vlm_model,
316
+ device = self.vlm_device,
317
+ load_4bit = self.vlm_load_4bit,
318
+ load_8bit = self.vlm_load_8bit
319
+ )
320
+
321
  # Create and clear the temporary directory.
322
  if not os.access('temp', os.F_OK):
323
  os.makedirs('temp')