#8900 Traceback (most recent call last): File

206.14**3 Posted at: 7 hours ago

Traceback (most recent call last):
File "I:\dy\dy\pyvideotrans\videotrans\mainwin\_actions_shot_match.py", line 173, in _run_shot_match

results = run_match_job(

File "I:\dy\dy\pyvideotrans\videotrans\task\shot_matcher.py", line 3831, in run_match_job

results = match_shots(

File "I:\dy\dy\pyvideotrans\videotrans\task\shot_matcher.py", line 3506, in match_shots

source_shots, indexed_source_samples, source_matrix = build_interval_source_feature_matrix(

File "I:\dy\dy\pyvideotrans\videotrans\task\shot_matcher.py", line 2635, in build_interval_source_feature_matrix

indexed_samples, matrix = build_visual_feature_matrix(samples, **feature_options)

File "I:\dy\dy\pyvideotrans\videotrans\task\shot_match_features.py", line 562, in build_visual_feature_matrix

return indexed_samples, feature_extractor.extract_batch(indexed_samples, batch_size=batch_size)

File "I:\dy\dy\pyvideotrans\videotrans\task\shot_match_features.py", line 489, in extract_batch

features.append(self.infer_prepared(prepared, batch_size=batch_size))

File "I:\dy\dy\pyvideotrans\videotrans\task\shot_match_features.py", line 441, in infer_prepared

features.append(self._forward(prepared[cursor:cursor+size]).cpu().numpy().astype(np.float32))

File "I:\dy\dy\pyvideotrans\videotrans\task\shot_match_features.py", line 333, in _forward

batch = self._to_device(prepared)

File "I:\dy\dy\pyvideotrans\videotrans\task\shot_match_features.py", line 329, in _to_device

return batch.to(self.device, non_blocking=True)

RuntimeError: CUDA error: unknown error
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1
Compile with TORCH_USE_CUDA_DSA to enable device-side assertions.


=
system:Windows-10-10.0.26100-SP0
version:v4.15
frozen:False
language:zh_CN
root_dir:I:/dy/dy/pyvideotrans
Python: 3.10.19 (main, Feb 12 2026, 00:37:08) [MSC v.1944 64 bit (AMD64)]

1 Replies

#1 DeepSeek reply at 7 hours ago

这是 CUDA 层报错:在“镜头匹配”提取视觉特征时,把张量传入 GPU 失败(batch.to(self.device))。常见触发原因有三类:显存不足、显卡驱动与 CUDA 版本不匹配,或上一次 GPU 运算崩溃后显存状态已损坏。该功能不在官方文档覆盖范围内,以下按 CUDA 通用原因排查。

  1. 重启软件后重试:CUDA error: unknown error 往往是上一次 GPU 内核崩溃后的残留状态;重启可重建 CUDA 上下文,有时可立即恢复。
  2. 降低批大小:在镜头匹配参数中调小 batch size(显存不足时最有效)。
  3. 改用 CPU 运行:若该功能提供设备选项,切换为 CPU,先确认是否是显存或驱动问题。
  4. 关闭其他占用显存的程序(浏览器硬件加速、其他 AI 任务),并将 高级选项-通用设置-GPU同时任务数 设为 1。
  5. 更新显卡驱动到最新版,确认 CUDA 12.8 + cuDNN 9.x 与所选显卡匹配。
  6. 需要定位具体出错点时,可设置环境变量 CUDA_LAUNCH_BLOCKING=1 后复现,以获得准确的报错 API。

若以上均无效,请提供软件目录下 logs/ 中最新的日志文件底部约 30 行信息(含显存大小、显卡型号),以便进一步判断。

Thinking...
Post Your Reply
If you have more than 50 messages, you will no longer receive an automatic reply. Please ask a new question!
Open source and free maintenance is not easy. If this project is helpful to you, please consider making a small donation to help the project continue to maintain and update.

Dondate via ko-fi.com

Related resource