Spaces:
Running
on
Zero
Running
on
Zero
FIXED RuntimeError: CUDA has been initialized before importing the `spaces` package
Browse files
app.py
CHANGED
@@ -1,8 +1,10 @@
|
|
|
|
1 |
import gradio as gr
|
2 |
import torch
|
3 |
torch.jit.script = lambda f: f # Avoid script error in lambda
|
4 |
from t2v_metrics import VQAScore, list_all_vqascore_models
|
5 |
|
|
|
6 |
def update_model(model_name):
|
7 |
return VQAScore(model=model_name, device="cuda")
|
8 |
|
@@ -11,13 +13,19 @@ global model_pipe, cur_model_name
|
|
11 |
cur_model_name = "clip-flant5-xl"
|
12 |
model_pipe = update_model(cur_model_name)
|
13 |
|
14 |
-
# Ensure GPU context manager is imported correctly (assuming spaces is a module you have)
|
15 |
-
try:
|
16 |
-
from spaces import GPU
|
17 |
-
except ImportError:
|
18 |
-
GPU = lambda duration: (lambda f: f) # Dummy decorator if spaces.GPU is not available
|
19 |
|
20 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
def generate(model_name, image, text):
|
22 |
global model_pipe, cur_model_name
|
23 |
|
@@ -38,7 +46,7 @@ def generate(model_name, image, text):
|
|
38 |
|
39 |
return result
|
40 |
|
41 |
-
|
42 |
def rank_images(model_name, images, text):
|
43 |
global model_pipe, cur_model_name
|
44 |
|
|
|
1 |
+
import spaces
|
2 |
import gradio as gr
|
3 |
import torch
|
4 |
torch.jit.script = lambda f: f # Avoid script error in lambda
|
5 |
from t2v_metrics import VQAScore, list_all_vqascore_models
|
6 |
|
7 |
+
|
8 |
def update_model(model_name):
|
9 |
return VQAScore(model=model_name, device="cuda")
|
10 |
|
|
|
13 |
cur_model_name = "clip-flant5-xl"
|
14 |
model_pipe = update_model(cur_model_name)
|
15 |
|
|
|
|
|
|
|
|
|
|
|
16 |
|
17 |
+
# Ensure GPU context manager is imported correctly (assuming spaces is a module you have)
|
18 |
+
#try:
|
19 |
+
#from spaces import GPU # i believe this is wrong, spaces package does not have "GPU"
|
20 |
+
#except ImportError:
|
21 |
+
# GPU = lambda duration: (lambda f: f) # Dummy decorator if spaces.GPU is not available
|
22 |
+
|
23 |
+
if torch.cuda.is_available():
|
24 |
+
model_pipe.device = "cuda"
|
25 |
+
else:
|
26 |
+
print("CUDA is not available")
|
27 |
+
|
28 |
+
@spaces.GPU # a duration lower than 60 does not work, leave as is.
|
29 |
def generate(model_name, image, text):
|
30 |
global model_pipe, cur_model_name
|
31 |
|
|
|
46 |
|
47 |
return result
|
48 |
|
49 |
+
|
50 |
def rank_images(model_name, images, text):
|
51 |
global model_pipe, cur_model_name
|
52 |
|