Spaces:
Running
on
Zero
Running
on
Zero
workaround for torch.jit
Browse files- demo/mesh_recon.py +6 -1
demo/mesh_recon.py
CHANGED
@@ -4,10 +4,14 @@ import numpy as np
|
|
4 |
import torch
|
5 |
import trimesh
|
6 |
|
|
|
7 |
|
8 |
-
|
|
|
9 |
|
10 |
# use torch hub
|
|
|
|
|
11 |
model = torch.hub.load("isl-org/ZoeDepth", "ZoeD_NK", pretrained=True).to(device).eval()
|
12 |
|
13 |
|
@@ -97,6 +101,7 @@ def depth_edges_mask(depth):
|
|
97 |
return mask
|
98 |
|
99 |
|
|
|
100 |
def mesh_reconstruction(
|
101 |
masked_image: np.ndarray,
|
102 |
mask: np.ndarray,
|
|
|
4 |
import torch
|
5 |
import trimesh
|
6 |
|
7 |
+
import spaces
|
8 |
|
9 |
+
|
10 |
+
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
|
11 |
|
12 |
# use torch hub
|
13 |
+
# zeroGPU hack from https://huggingface.co/spaces/zero-gpu-explorers/README/discussions/9
|
14 |
+
torch.jit.script = lambda f: f
|
15 |
model = torch.hub.load("isl-org/ZoeDepth", "ZoeD_NK", pretrained=True).to(device).eval()
|
16 |
|
17 |
|
|
|
101 |
return mask
|
102 |
|
103 |
|
104 |
+
@spaces.GPU
|
105 |
def mesh_reconstruction(
|
106 |
masked_image: np.ndarray,
|
107 |
mask: np.ndarray,
|