双显卡，核显OK，独显报错

我电脑核显+独显4070.核显完全正常，选择独显会报错如下，提示MSVC未发现，但是我安装了2022版本的全套，不然核显执也是不能用， 独显日志显式torch.cuda.is_available() is False,我直接在latensync模型下的虚拟环境执行python也是true返回。并不是False。并且核显完全正常使用，也显式GPU加速。所以感觉很大概率是项目BUG，或者是latensync这个项目不支持我的显卡cuda?我显卡cuda是12.2的。

0

time="2025-03-25T18:23:14+08:00" level=info msg="server starting" buildTime="2025-03-05 09:48:07" debug=false version=0.0.1

1

time="2025-03-25T18:23:15+08:00" level=info msg="start command mode"

2

################## AigcPanelServer ###################

3

root: D:\AigcPanel\models\aigcpanel-server-latentsync-win-x86-v0.1.1

4

platformName: win

5

platformArch: x86

6

name: latentsync

7

######################################################

8

2025-03-25 18:23:15 - 开始运行

9

2025-03-25 18:23:15 - config {'id': 'VideoGen_20', 'mode': 'local', 'modelConfig': {'type': 'videoGen', 'param': {}, 'video': 'C:\\Users\\guojian\\AppData\\Roaming\\aigcpanel\\data\\videoTemplate\\wadrZCEzHBbhJbnr.mp4', 'audio': 'C:\\Users\\guojian\\AppData\\Roaming\\aigcpanel\\data\\soundTts\\1.wav'}, 'setting': {'gpu': '1'}} ['D:\\AigcPanel\\models\\aigcpanel-server-latentsync-win-x86-v0.1.1\\aigcpanelrun.py', 'C:\\Users\\guojian\\AppData\\Roaming\\aigcpanel\\temp\\file_1742898194199_D0zeUYteRdz9w55cPJi6Ca97MQVvRN42.json']

10

2025-03-25 18:23:15 - set CUDA_VISIBLE_DEVICES 1

11

2025-03-25 18:23:18 - 开始运行 {'UseCuda': False}

12

Result[VideoGen_20][{"UseCuda": false}]

13

AigcPanelRunResult[VideoGen_20][eyJVc2VDdWRhIjogZmFsc2V9]

14

Traceback (most recent call last):

15

File "D:\AigcPanel\models\aigcpanel-server-latentsync-win-x86-v0.1.1\_aienv\lib\runpy.py", line 196, in _run_module_as_main

16

return _run_code(code, main_globals, None,

17

File "D:\AigcPanel\models\aigcpanel-server-latentsync-win-x86-v0.1.1\_aienv\lib\runpy.py", line 86, in _run_code

18

exec(code, run_globals)

19

File "D:\AigcPanel\models\aigcpanel-server-latentsync-win-x86-v0.1.1\scripts\inference.py", line 105, in <module>

20

main(config, args)

21

File "D:\AigcPanel\models\aigcpanel-server-latentsync-win-x86-v0.1.1\scripts\inference.py", line 44, in main

22

audio_encoder = Audio2Feature(model_path=whisper_model_path, device="cuda", num_frames=config.data.num_frames)

23

File "D:\AigcPanel\models\aigcpanel-server-latentsync-win-x86-v0.1.1\latentsync\whisper\audio2feature.py", line 17, in __init__

24

self.model = load_model(model_path, device)

25

File "D:\AigcPanel\models\aigcpanel-server-latentsync-win-x86-v0.1.1\latentsync\whisper\whisper\__init__.py", line 112, in load_model

26

checkpoint = torch.load(fp, map_location=device)

27

File "D:\AigcPanel\models\aigcpanel-server-latentsync-win-x86-v0.1.1\_aienv\lib\site-packages\torch\serialization.py", line 1360, in load

28

return _load(

29

File "D:\AigcPanel\models\aigcpanel-server-latentsync-win-x86-v0.1.1\_aienv\lib\site-packages\torch\serialization.py", line 1848, in _load

30

result = unpickler.load()

31

File "D:\AigcPanel\models\aigcpanel-server-latentsync-win-x86-v0.1.1\_aienv\lib\site-packages\torch\serialization.py", line 1812, in persistent_load

32

typed_storage = load_tensor(

33

File "D:\AigcPanel\models\aigcpanel-server-latentsync-win-x86-v0.1.1\_aienv\lib\site-packages\torch\serialization.py", line 1784, in load_tensor

34

wrap_storage=restore_location(storage, location),

35

File "D:\AigcPanel\models\aigcpanel-server-latentsync-win-x86-v0.1.1\_aienv\lib\site-packages\torch\serialization.py", line 1685, in restore_location

36

return default_restore_location(storage, map_location)

37

File "D:\AigcPanel\models\aigcpanel-server-latentsync-win-x86-v0.1.1\_aienv\lib\site-packages\torch\serialization.py", line 601, in default_restore_location

38

result = fn(storage, location)

39

File "D:\AigcPanel\models\aigcpanel-server-latentsync-win-x86-v0.1.1\_aienv\lib\site-packages\torch\serialization.py", line 539, in _deserialize

40

device = _validate_device(location, backend_name)

41

File "D:\AigcPanel\models\aigcpanel-server-latentsync-win-x86-v0.1.1\_aienv\lib\site-packages\torch\serialization.py", line 508, in _validate_device

42

raise RuntimeError(

43

RuntimeError: Attempting to deserialize object on a CUDA device but torch.cuda.is_available() is False. If you are running on a CPU-only machine, please use torch.load with map_location=torch.device('cpu') to map your storages to the CPU.

44

WARNING: Failed to find MSVC.

45

Input video path: C:\Users\guojian\AppData\Roaming\aigcpanel\data\videoTemplate\wadrZCEzHBbhJbnr.mp4

46

Input audio path: C:\Users\guojian\AppData\Roaming\aigcpanel\data\soundTts\1.wav

47

Loaded checkpoint path: aigcpanelmodels/checkpoints/latentsync_unet.pt

48

Error: run D:\AigcPanel\models\aigcpanel-server-latentsync-win-x86-v0.1.1\_aienv\python.exe -m scripts.inference --unet_config_path configs/unet/second_stage.yaml --inference_ckpt_path aigcpanelmodels/checkpoints/latentsync_unet.pt --inference_steps 20 --guidance_scale 1.5 --video_path C:\Users\guojian\AppData\Roaming\aigcpanel\data\videoTemplate\wadrZCEzHBbhJbnr.mp4 --audio_path C:\Users\guojian\AppData\Roaming\aigcpanel\data\soundTts\1.wav --video_out_path D:\AigcPanel\models\aigcpanel-server-latentsync-win-x86-v0.1.1\_cache/file\e582ab93de5a1a8b382c42d929978035.mp4 failed: CompletedProcess(args=['D:\\AigcPanel\\models\\aigcpanel-server-latentsync-win-x86-v0.1.1\\_aienv\\python.exe', '-m', 'scripts.inference', '--unet_config_path', 'configs/unet/second_stage.yaml', '--inference_ckpt_path', 'aigcpanelmodels/checkpoints/latentsync_unet.pt', '--inference_steps', '20', '--guidance_scale', '1.5', '--video_path', 'C:\\Users\\guojian\\AppData\\Roaming\\aigcpanel\\data\\videoTemplate\\wadrZCEzHBbhJbnr.mp4', '--audio_path', 'C:\\Users\\guojian\\AppData\\Roaming\\aigcpanel\\data\\soundTts\\1.wav', '--video_out_path', 'D:\\AigcPanel\\models\\aigcpanel-server-latentsync-win-x86-v0.1.1\\_cache/file\\e582ab93de5a1a8b382c42d929978035.mp4'], returncode=1)

49

time="2025-03-25T18:23:22+08:00" level=error msg="Error:exit status 0xffffffff"