Fred Zhang
commited on
Commit
•
eaacc51
1
Parent(s):
3826cd5
add model
Browse files- Config.py +36 -0
- Model.py +117 -0
- config.json +48 -0
- pytorch_model.bin +3 -0
Config.py
ADDED
@@ -0,0 +1,36 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from transformers import PretrainedConfig
|
2 |
+
from typing import List
|
3 |
+
|
4 |
+
classes_example = {
|
5 |
+
0: 'nsfw_gore',
|
6 |
+
1: 'nsfw_suggestive',
|
7 |
+
2: 'safe'
|
8 |
+
}
|
9 |
+
|
10 |
+
class InceptionV3Config(PretrainedConfig):
|
11 |
+
model_type = "inceptionv3"
|
12 |
+
def __init__(self, model_name: str = "inception_v3", input_channels: int = 3, num_classes: int = 3, input_size: List[int] = [3, 299, 299], pool_size: List[int] = [8, 8, 2048], crop_pct: float = 0.875, interpolation: str = "bicubic", mean: List[float] = [0.5, 0.5, 0.5], std: List[float] = [0.5, 0.5, 0.5], first_conv: str = "Conv2d_1a_3x3.conv", classifier: str = "fc", has_aux: bool = True, label_offset: int = 1, classes: dict = classes_example, output_channels: int = 2048, use_jit=False, **kwargs):
|
13 |
+
self.model_name = model_name
|
14 |
+
self.input_channels = input_channels
|
15 |
+
self.num_classes = num_classes
|
16 |
+
self.input_size = input_size
|
17 |
+
self.pool_size = pool_size
|
18 |
+
self.crop_pct = crop_pct
|
19 |
+
self.interpolation = interpolation
|
20 |
+
self.mean = mean
|
21 |
+
self.std = std
|
22 |
+
self.first_conv = first_conv
|
23 |
+
self.classifier = classifier
|
24 |
+
self.has_aux = has_aux
|
25 |
+
self.label_offset = label_offset
|
26 |
+
self.classes = classes
|
27 |
+
self.output_channels = output_channels
|
28 |
+
self.use_jit = use_jit
|
29 |
+
super().__init__(**kwargs)
|
30 |
+
|
31 |
+
"""
|
32 |
+
|
33 |
+
inceptionv3_config = InceptionV3Config()
|
34 |
+
inceptionv3_config.save_pretrained("inceptionv3_config")
|
35 |
+
|
36 |
+
"""
|
Model.py
ADDED
@@ -0,0 +1,117 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from transformers import PreTrainedModel
|
2 |
+
import torch
|
3 |
+
import os
|
4 |
+
|
5 |
+
url_map = {
|
6 |
+
"inception_v3": "https://download.pytorch.org/models/inception_v3_google-1a9a5a14.pth"
|
7 |
+
}
|
8 |
+
|
9 |
+
class InceptionV3ModelForImageClassification(PreTrainedModel):
|
10 |
+
def __init__(self, config):
|
11 |
+
super().__init__(config)
|
12 |
+
|
13 |
+
model_path = f"{self.config.model_name}.bin".replace("/","_")
|
14 |
+
|
15 |
+
if self.config.model_name == "google-safesearch-mini":
|
16 |
+
self.model = torch.jit.load(model_path)
|
17 |
+
elif self.config.model_name == "inception_v3":
|
18 |
+
self.model = torch.hub.load('pytorch/vision:v0.6.0', 'inception_v3', pretrained=True)
|
19 |
+
else:
|
20 |
+
if not os.path.exists(model_path):
|
21 |
+
from urllib.request import urlretrieve
|
22 |
+
urlretrieve(f"https://huggingface.co/{self.config.model_name}/resolve/main/pytorch_model.bin", model_path)
|
23 |
+
self.model = torch.jit.load(model_path) if self.config.use_jit else torch.load(model_path)
|
24 |
+
|
25 |
+
def forward(self, input_ids):
|
26 |
+
out, aux = self.model(input_ids)
|
27 |
+
return out, aux
|
28 |
+
|
29 |
+
def freeze(self):
|
30 |
+
for param in self.model.parameters():
|
31 |
+
param.requires_grad = False
|
32 |
+
|
33 |
+
def unfreeze(self):
|
34 |
+
for param in self.model.parameters():
|
35 |
+
param.requires_grad = True
|
36 |
+
|
37 |
+
def train(self, mode=True):
|
38 |
+
super().train(mode)
|
39 |
+
self.model.train(mode)
|
40 |
+
|
41 |
+
def eval(self):
|
42 |
+
return self.train(False)
|
43 |
+
|
44 |
+
def to(self, device):
|
45 |
+
self.model.to(device)
|
46 |
+
return self
|
47 |
+
|
48 |
+
def cuda(self, device=None):
|
49 |
+
return self.to("cuda")
|
50 |
+
|
51 |
+
def cpu(self):
|
52 |
+
return self.to("cpu")
|
53 |
+
|
54 |
+
def state_dict(self, destination=None, prefix='', keep_vars=False):
|
55 |
+
return self.model.state_dict(destination, prefix, keep_vars)
|
56 |
+
|
57 |
+
def load_state_dict(self, state_dict, strict=True):
|
58 |
+
return self.model.load_state_dict(state_dict, strict)
|
59 |
+
|
60 |
+
def parameters(self, recurse=True):
|
61 |
+
return self.model.parameters(recurse)
|
62 |
+
|
63 |
+
def named_parameters(self, prefix='', recurse=True):
|
64 |
+
return self.model.named_parameters(prefix, recurse)
|
65 |
+
|
66 |
+
def children(self):
|
67 |
+
return self.model.children()
|
68 |
+
|
69 |
+
def named_children(self):
|
70 |
+
return self.model.named_children()
|
71 |
+
|
72 |
+
def modules(self):
|
73 |
+
return self.model.modules()
|
74 |
+
|
75 |
+
def named_modules(self, memo=None, prefix=''):
|
76 |
+
return self.model.named_modules(memo, prefix)
|
77 |
+
|
78 |
+
def zero_grad(self, set_to_none=False):
|
79 |
+
return self.model.zero_grad(set_to_none)
|
80 |
+
|
81 |
+
def share_memory(self):
|
82 |
+
return self.model.share_memory()
|
83 |
+
|
84 |
+
def transform(self, image):
|
85 |
+
from torchvision import transforms
|
86 |
+
transform = transforms.Compose([
|
87 |
+
transforms.Resize(299),
|
88 |
+
transforms.ToTensor(),
|
89 |
+
transforms.Normalize(mean=self.config.mean, std=self.config.std)
|
90 |
+
])
|
91 |
+
image = transform(image)
|
92 |
+
return image
|
93 |
+
|
94 |
+
def open_image(self, path):
|
95 |
+
from PIL import Image
|
96 |
+
path = 'https://images.unsplash.com/photo-1594568284297-7c64464062b1'
|
97 |
+
if path.startswith('http://') or path.startswith('https://'):
|
98 |
+
import requests
|
99 |
+
from io import BytesIO
|
100 |
+
response = requests.get(path)
|
101 |
+
image = Image.open(BytesIO(response.content)).convert('RGB')
|
102 |
+
else:
|
103 |
+
image = Image.open(path).convert('RGB')
|
104 |
+
return image
|
105 |
+
|
106 |
+
def predict(self, path, device="cuda"):
|
107 |
+
image = self.open_image(path)
|
108 |
+
image = self.transform(image)
|
109 |
+
image = image.unsqueeze(0)
|
110 |
+
self.eval()
|
111 |
+
if device == "cuda":
|
112 |
+
image = image.cuda()
|
113 |
+
with torch.no_grad():
|
114 |
+
out, aux = self(image)
|
115 |
+
print(out)
|
116 |
+
_, predicted = torch.max(out.data, 1)
|
117 |
+
return self.config.classes[predicted.item()]
|
config.json
ADDED
@@ -0,0 +1,48 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"architectures": [
|
3 |
+
"InceptionV3ModelForImageClassification"
|
4 |
+
],
|
5 |
+
"auto_map": {
|
6 |
+
"AutoConfig": "Config.InceptionV3Config",
|
7 |
+
"AutoModelForImageClassification": "Model.InceptionV3ModelForImageClassification"
|
8 |
+
},
|
9 |
+
"classes": {
|
10 |
+
"0": "nsfw_gore",
|
11 |
+
"1": "nsfw_suggestive",
|
12 |
+
"2": "safe"
|
13 |
+
},
|
14 |
+
"classifier": "fc",
|
15 |
+
"crop_pct": 0.875,
|
16 |
+
"first_conv": "Conv2d_1a_3x3.conv",
|
17 |
+
"has_aux": true,
|
18 |
+
"input_channels": 3,
|
19 |
+
"input_size": [
|
20 |
+
3,
|
21 |
+
299,
|
22 |
+
299
|
23 |
+
],
|
24 |
+
"interpolation": "bicubic",
|
25 |
+
"label_offset": 1,
|
26 |
+
"mean": [
|
27 |
+
0.5,
|
28 |
+
0.5,
|
29 |
+
0.5
|
30 |
+
],
|
31 |
+
"model_name": "google-safesearch-mini",
|
32 |
+
"model_type": "inceptionv3",
|
33 |
+
"num_classes": 3,
|
34 |
+
"output_channels": 2048,
|
35 |
+
"pool_size": [
|
36 |
+
8,
|
37 |
+
8,
|
38 |
+
2048
|
39 |
+
],
|
40 |
+
"std": [
|
41 |
+
0.5,
|
42 |
+
0.5,
|
43 |
+
0.5
|
44 |
+
],
|
45 |
+
"torch_dtype": "float32",
|
46 |
+
"transformers_version": "4.21.2",
|
47 |
+
"use_jit": true
|
48 |
+
}
|
pytorch_model.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:db510376e428b0d5f1472e4f56d31a4bfbee69b3e8a58c67a802098e00d42d12
|
3 |
+
size 100804217
|