Spaces:
Running
on
Zero
Running
on
Zero
NIRVANALAN
commited on
Commit
β’
89337c5
1
Parent(s):
a0afc03
update dep
Browse files- README.md +1 -2
- app.py +1 -1
- dit/dit_models_xformers.py +2 -2
- ldm/modules/attention.py +4 -4
README.md
CHANGED
@@ -3,8 +3,7 @@ title: LN3DIff-I23D
|
|
3 |
emoji: π
|
4 |
colorFrom: green
|
5 |
colorTo: green
|
6 |
-
|
7 |
-
sdk: docker
|
8 |
sdk_version: 4.25.0
|
9 |
app_file: app.py
|
10 |
pinned: false
|
|
|
3 |
emoji: π
|
4 |
colorFrom: green
|
5 |
colorTo: green
|
6 |
+
sdk: gradio
|
|
|
7 |
sdk_version: 4.25.0
|
8 |
app_file: app.py
|
9 |
pinned: false
|
app.py
CHANGED
@@ -49,7 +49,7 @@ th.backends.cuda.matmul.allow_tf32 = True
|
|
49 |
th.backends.cudnn.allow_tf32 = True
|
50 |
th.backends.cudnn.enabled = True
|
51 |
|
52 |
-
install_dependency()
|
53 |
|
54 |
from guided_diffusion import dist_util, logger
|
55 |
from guided_diffusion.script_util import (
|
|
|
49 |
th.backends.cudnn.allow_tf32 = True
|
50 |
th.backends.cudnn.enabled = True
|
51 |
|
52 |
+
# install_dependency()
|
53 |
|
54 |
from guided_diffusion import dist_util, logger
|
55 |
from guided_diffusion.script_util import (
|
dit/dit_models_xformers.py
CHANGED
@@ -28,8 +28,8 @@ try:
|
|
28 |
from apex.normalization import FusedRMSNorm as RMSNorm
|
29 |
from apex.normalization import FusedLayerNorm as LayerNorm
|
30 |
except:
|
31 |
-
from torch.nn import LayerNorm
|
32 |
-
from .
|
33 |
|
34 |
# from torch.nn import LayerNorm
|
35 |
# from xformers import triton
|
|
|
28 |
from apex.normalization import FusedRMSNorm as RMSNorm
|
29 |
from apex.normalization import FusedLayerNorm as LayerNorm
|
30 |
except:
|
31 |
+
from torch.nn import LayerNorm
|
32 |
+
from torch.nn import RMSNorm # requires torch2.4
|
33 |
|
34 |
# from torch.nn import LayerNorm
|
35 |
# from xformers import triton
|
ldm/modules/attention.py
CHANGED
@@ -14,10 +14,10 @@ from ldm.modules.diffusionmodules.util import checkpoint
|
|
14 |
import os
|
15 |
_ATTN_PRECISION = os.environ.get("ATTN_PRECISION", "fp32")
|
16 |
from xformers.ops import MemoryEfficientAttentionFlashAttentionOp
|
17 |
-
|
18 |
-
from apex.normalization import FusedRMSNorm as RMSNorm
|
19 |
-
|
20 |
-
|
21 |
|
22 |
|
23 |
def exists(val):
|
|
|
14 |
import os
|
15 |
_ATTN_PRECISION = os.environ.get("ATTN_PRECISION", "fp32")
|
16 |
from xformers.ops import MemoryEfficientAttentionFlashAttentionOp
|
17 |
+
try:
|
18 |
+
from apex.normalization import FusedRMSNorm as RMSNorm
|
19 |
+
except:
|
20 |
+
from torch.nn import RMSNorm # requires torch2.4
|
21 |
|
22 |
|
23 |
def exists(val):
|