how to

pytorch 奇怪报错记录 for grep

Jun 17, 2025
notesjulyfun2506
2 Minutes
357 Words

pytorch 奇怪报错记录 for grep

25.6.17

1
---------------------------------------------------------------------------
2
RuntimeError Traceback (most recent call last)
3
File /home/julyfun/Documents/GitHub/julyfun/how-to/notes/julyfun/技术学习/diffusion-models-class/unit2-02_class_conditioned_diffusion_model_example.py:3
4
1 # %%
5
2 @run
6
----> 3 def func():
7
4 net = ClassConditionedUnet(num_classes=10).to(device)
8
5 img = torch.randn(3, 1, 28, 28).to(device)
9
10
File ~/Documents/GitHub/julyfun/robotoy/robotoy/ziglike/test.py:58, in run(func)
11
57 def run(func):
12
---> 58 func()
13
14
File /home/julyfun/Documents/GitHub/julyfun/how-to/notes/julyfun/技术学习/diffusion-models-class/unit2-02_class_conditioned_diffusion_model_example.py:4
15
2 @run
51 collapsed lines
16
3 def func():
17
----> 4 net = ClassConditionedUnet(num_classes=10).to(device)
18
5 img = torch.randn(3, 1, 28, 28).to(device)
19
6 cls = torch.tensor([0, 1, 9]).to(device)
20
21
File ~/Documents/GitHub/diffusion-models-class/.venv/lib/python3.10/site-packages/torch/nn/modules/module.py:1355, in Module.to(self, *args, **kwargs)
22
1352 else:
23
1353 raise
24
-> 1355 return self._apply(convert)
25
26
File ~/Documents/GitHub/diffusion-models-class/.venv/lib/python3.10/site-packages/torch/nn/modules/module.py:915, in Module._apply(self, fn, recurse)
27
913 if recurse:
28
914 for module in self.children():
29
--> 915 module._apply(fn)
30
917 def compute_should_use_set_data(tensor, tensor_applied):
31
918 if torch._has_compatible_shallow_copy_type(tensor, tensor_applied):
32
919 # If the new tensor has compatible tensor type as the existing tensor,
33
920 # the current behavior is to change the tensor in-place using `.data =`,
34
(...)
35
925 # global flag to let the user control whether they want the future
36
926 # behavior of overwriting the existing tensor or not.
37
38
File ~/Documents/GitHub/diffusion-models-class/.venv/lib/python3.10/site-packages/torch/nn/modules/module.py:942, in Module._apply(self, fn, recurse)
39
938 # Tensors stored in modules are graph leaves, and we don't want to
40
939 # track autograd history of `param_applied`, so we have to use
41
940 # `with torch.no_grad():`
42
941 with torch.no_grad():
43
--> 942 param_applied = fn(param)
44
943 p_should_use_set_data = compute_should_use_set_data(param, param_applied)
45
945 # subclasses may have multiple child tensors so we need to use swap_tensors
46
47
File ~/Documents/GitHub/diffusion-models-class/.venv/lib/python3.10/site-packages/torch/nn/modules/module.py:1341, in Module.to.<locals>.convert(t)
48
1334 if convert_to_format is not None and t.dim() in (4, 5):
49
1335 return t.to(
50
1336 device,
51
1337 dtype if t.is_floating_point() or t.is_complex() else None,
52
1338 non_blocking,
53
1339 memory_format=convert_to_format,
54
1340 )
55
-> 1341 return t.to(
56
1342 device,
57
1343 dtype if t.is_floating_point() or t.is_complex() else None,
58
1344 non_blocking,
59
1345 )
60
1346 except NotImplementedError as e:
61
1347 if str(e) == "Cannot copy out of meta tensor; no data!":
62
63
RuntimeError: CUDA error: device-side assert triggered
64
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
65
For debugging consider passing CUDA_LAUNCH_BLOCKING=1
66
Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.
  • context see above.
  • [ok, but why] 重启 kernel 运行就没问题了.
Article title:pytorch 奇怪报错记录 for grep
Article author:Julyfun
Release time:Jun 17, 2025
Copyright 2025
Sitemap