xzerus commited on
Commit
1744947
·
verified ·
1 Parent(s): f511cdd

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +5 -6
app.py CHANGED
@@ -6,17 +6,17 @@ from torchvision.transforms.functional import InterpolationMode
6
  from transformers import AutoModel, AutoTokenizer
7
  import gradio as gr
8
 
 
9
  IMAGENET_MEAN = (0.485, 0.456, 0.406)
10
  IMAGENET_STD = (0.229, 0.224, 0.225)
11
 
12
  # Build the image transform
13
  def build_transform(input_size):
14
- MEAN, STD = IMAGENET_MEAN, IMAGENET_STD
15
  transform = T.Compose([
16
  T.Lambda(lambda img: img.convert('RGB') if img.mode != 'RGB' else img),
17
  T.Resize((input_size, input_size), interpolation=InterpolationMode.BICUBIC),
18
  T.ToTensor(),
19
- T.Normalize(mean=MEAN, std=STD)
20
  ])
21
  return transform
22
 
@@ -60,10 +60,9 @@ path = 'OpenGVLab/InternVL2_5-78B'
60
  model = AutoModel.from_pretrained(
61
  path,
62
  torch_dtype=torch.bfloat16,
63
- low_cpu_mem_usage=True,
64
- use_flash_attn=True,
65
- trust_remote_code=True
66
- ).eval().cuda()
67
 
68
  tokenizer = AutoTokenizer.from_pretrained(path, trust_remote_code=True, use_fast=False)
69
 
 
6
  from transformers import AutoModel, AutoTokenizer
7
  import gradio as gr
8
 
9
+ # Constants for ImageNet preprocessing
10
  IMAGENET_MEAN = (0.485, 0.456, 0.406)
11
  IMAGENET_STD = (0.229, 0.224, 0.225)
12
 
13
  # Build the image transform
14
  def build_transform(input_size):
 
15
  transform = T.Compose([
16
  T.Lambda(lambda img: img.convert('RGB') if img.mode != 'RGB' else img),
17
  T.Resize((input_size, input_size), interpolation=InterpolationMode.BICUBIC),
18
  T.ToTensor(),
19
+ T.Normalize(mean=IMAGENET_MEAN, std=IMAGENET_STD)
20
  ])
21
  return transform
22
 
 
60
  model = AutoModel.from_pretrained(
61
  path,
62
  torch_dtype=torch.bfloat16,
63
+ trust_remote_code=True,
64
+ device_map="auto" # Use device map for efficient memory handling
65
+ )
 
66
 
67
  tokenizer = AutoTokenizer.from_pretrained(path, trust_remote_code=True, use_fast=False)
68