Spaces:
Running
on
Zero
Running
on
Zero
brandonsmart
commited on
Commit
•
c65a25a
1
Parent(s):
e059ab1
Attempting to solve pickle issue
Browse files
demo.py
CHANGED
@@ -23,8 +23,10 @@ from mast3r.utils.misc import hash_md5
|
|
23 |
import main
|
24 |
import utils.export as export
|
25 |
|
26 |
-
@spaces.GPU(duration=
|
27 |
-
def get_reconstructed_scene(outdir, model,
|
|
|
|
|
28 |
|
29 |
assert len(filelist) == 1 or len(filelist) == 2, "Please provide one or two images"
|
30 |
if ios_mode:
|
@@ -37,6 +39,7 @@ def get_reconstructed_scene(outdir, model, device, silent, image_size, ios_mode,
|
|
37 |
img['img'] = img['img'].to(device)
|
38 |
img['original_img'] = img['original_img'].to(device)
|
39 |
img['true_shape'] = torch.from_numpy(img['true_shape'])
|
|
|
40 |
|
41 |
output = model(imgs[0], imgs[1])
|
42 |
|
@@ -50,13 +53,11 @@ if __name__ == '__main__':
|
|
50 |
image_size = 512
|
51 |
silent = False
|
52 |
ios_mode = True
|
53 |
-
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
|
54 |
|
55 |
model_name = "brandonsmart/splatt3r_v1.0"
|
56 |
filename = "epoch=19-step=1200.ckpt"
|
57 |
weights_path = hf_hub_download(repo_id=model_name, filename=filename)
|
58 |
model = main.MAST3RGaussians.load_from_checkpoint(weights_path, 'cpu')
|
59 |
-
model = model.to(device)
|
60 |
chkpt_tag = hash_md5(weights_path)
|
61 |
|
62 |
# Define example inputs and their corresponding precalculated outputs
|
@@ -88,7 +89,7 @@ if __name__ == '__main__':
|
|
88 |
cache_path = os.path.join(tmpdirname, chkpt_tag)
|
89 |
os.makedirs(cache_path, exist_ok=True)
|
90 |
|
91 |
-
recon_fun = functools.partial(get_reconstructed_scene, tmpdirname, model,
|
92 |
|
93 |
if not ios_mode:
|
94 |
for i in range(len(examples)):
|
|
|
23 |
import main
|
24 |
import utils.export as export
|
25 |
|
26 |
+
@spaces.GPU(duration=15)
|
27 |
+
def get_reconstructed_scene(outdir, model, silent, image_size, ios_mode, filelist):
|
28 |
+
|
29 |
+
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
|
30 |
|
31 |
assert len(filelist) == 1 or len(filelist) == 2, "Please provide one or two images"
|
32 |
if ios_mode:
|
|
|
39 |
img['img'] = img['img'].to(device)
|
40 |
img['original_img'] = img['original_img'].to(device)
|
41 |
img['true_shape'] = torch.from_numpy(img['true_shape'])
|
42 |
+
model = model.to(device)
|
43 |
|
44 |
output = model(imgs[0], imgs[1])
|
45 |
|
|
|
53 |
image_size = 512
|
54 |
silent = False
|
55 |
ios_mode = True
|
|
|
56 |
|
57 |
model_name = "brandonsmart/splatt3r_v1.0"
|
58 |
filename = "epoch=19-step=1200.ckpt"
|
59 |
weights_path = hf_hub_download(repo_id=model_name, filename=filename)
|
60 |
model = main.MAST3RGaussians.load_from_checkpoint(weights_path, 'cpu')
|
|
|
61 |
chkpt_tag = hash_md5(weights_path)
|
62 |
|
63 |
# Define example inputs and their corresponding precalculated outputs
|
|
|
89 |
cache_path = os.path.join(tmpdirname, chkpt_tag)
|
90 |
os.makedirs(cache_path, exist_ok=True)
|
91 |
|
92 |
+
recon_fun = functools.partial(get_reconstructed_scene, tmpdirname, model, silent, image_size, ios_mode)
|
93 |
|
94 |
if not ios_mode:
|
95 |
for i in range(len(examples)):
|