Spaces:
Running
on
Zero
Running
on
Zero
fix: different safetensors
Browse files
app.py
CHANGED
@@ -5,11 +5,27 @@ from PIL import Image
|
|
5 |
from net.CIDNet import CIDNet
|
6 |
import torchvision.transforms as transforms
|
7 |
import torch.nn.functional as F
|
8 |
-
import
|
9 |
import imquality.brisque as brisque
|
10 |
from loss.niqe_utils import *
|
11 |
import spaces
|
12 |
-
import
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
13 |
|
14 |
eval_net = CIDNet().cuda()
|
15 |
eval_net.trans.gated = True
|
@@ -20,7 +36,7 @@ def process_image(input_img,score,model_path,gamma=1.0,alpha_s=1.0,alpha_i=1.0):
|
|
20 |
if model_path is None:
|
21 |
return input_img,"Please choose a model weights."
|
22 |
torch.set_grad_enabled(False)
|
23 |
-
|
24 |
# eval_net.load_state_dict(torch.load(os.path.join(directory,model_path), map_location=lambda storage, loc: storage))
|
25 |
eval_net.eval()
|
26 |
|
@@ -81,4 +97,4 @@ interface = gr.Interface(
|
|
81 |
allow_flagging="never"
|
82 |
)
|
83 |
|
84 |
-
interface.launch(share=
|
|
|
5 |
from net.CIDNet import CIDNet
|
6 |
import torchvision.transforms as transforms
|
7 |
import torch.nn.functional as F
|
8 |
+
import safetensors.torch as sf
|
9 |
import imquality.brisque as brisque
|
10 |
from loss.niqe_utils import *
|
11 |
import spaces
|
12 |
+
from huggingface_hub import hf_hub_download
|
13 |
+
import json
|
14 |
+
|
15 |
+
def from_pretrained(cls, pretrained_model_name_or_path: str):
|
16 |
+
model_id = str(pretrained_model_name_or_path)
|
17 |
+
|
18 |
+
config_file = hf_hub_download(repo_id=model_id, filename="config.json", repo_type="model")
|
19 |
+
config = None
|
20 |
+
if config_file is not None:
|
21 |
+
with open(config_file, "r", encoding="utf-8") as f:
|
22 |
+
config = json.load(f)
|
23 |
+
|
24 |
+
|
25 |
+
model_file = hf_hub_download(repo_id=model_id, filename="model.safetensors", repo_type="model")
|
26 |
+
# instance = sf.load_model(cls, model_file, strict=False)
|
27 |
+
state_dict = sf.load_file(model_file)
|
28 |
+
cls.load_state_dict(state_dict, strict=False)
|
29 |
|
30 |
eval_net = CIDNet().cuda()
|
31 |
eval_net.trans.gated = True
|
|
|
36 |
if model_path is None:
|
37 |
return input_img,"Please choose a model weights."
|
38 |
torch.set_grad_enabled(False)
|
39 |
+
from_pretrained(eval_net,"Fediory/HVI-CIDNet-"+model_path)
|
40 |
# eval_net.load_state_dict(torch.load(os.path.join(directory,model_path), map_location=lambda storage, loc: storage))
|
41 |
eval_net.eval()
|
42 |
|
|
|
97 |
allow_flagging="never"
|
98 |
)
|
99 |
|
100 |
+
interface.launch(share=False)
|