Spaces:
Running on Zero
Running on Zero
Commit ·
35c95b4
1
Parent(s): 21ece0c
Fix HunyuanFoley xregen AttributeError: return torch.device from _get_device_and_dtype
Browse filesdenoise_process calls torch.autocast(device_type=device.type, ...) which
requires a torch.device object. _get_device_and_dtype was returning a plain
string 'cuda', causing AttributeError: 'str' object has no attribute 'type'.
All other usages (.to(), map_location=, device=) work with torch.device too.
app.py
CHANGED
|
@@ -175,8 +175,9 @@ def _ensure_syspath(subdir: str) -> str:
|
|
| 175 |
|
| 176 |
|
| 177 |
def _get_device_and_dtype() -> tuple:
|
| 178 |
-
"""Return (device, weight_dtype) pair used by all GPU functions.
|
| 179 |
-
device
|
|
|
|
| 180 |
return device, torch.bfloat16
|
| 181 |
|
| 182 |
|
|
|
|
| 175 |
|
| 176 |
|
| 177 |
def _get_device_and_dtype() -> tuple:
|
| 178 |
+
"""Return (device, weight_dtype) pair used by all GPU functions.
|
| 179 |
+
Returns a torch.device object so callers can use device.type safely."""
|
| 180 |
+
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
| 181 |
return device, torch.bfloat16
|
| 182 |
|
| 183 |
|