#3681 RuntimeError: _process:Unable to allocate 1.38 GiB for an array with shape (1, 462395, 400) and data type float64

36.152* Posted at: 17 hours ago 👁20

语音识别阶段出错:_process:Unable to allocate 1.38 GiB for an array with shape (1, 462395, 400) and data type float64:Traceback (most recent call last):
File "videotrans\task\trans_create.py", line 376, in recogn
File "videotrans\recognition\__init__.py", line 250, in run
File "videotrans\recognition\_base.py", line 68, in run
File "videotrans\recognition\_overall.py", line 144, in _exec
RuntimeError: _process:Unable to allocate 1.38 GiB for an array with shape (1, 462395, 400) and data type float64

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "videotrans\task\job.py", line 79, in run
File "videotrans\task\trans_create.py", line 412, in recogn
RuntimeError: _process:Unable to allocate 1.38 GiB for an array with shape (1, 462395, 400) and data type float64


=
Windows-10-10.0.26200-SP0
version:v3.79
frozen:True
language:zh

1 Replies

这是由于显存或内存不足导致的。错误信息显示尝试分配 1.38 GiB 的数组失败,通常是因为使用了大型语音识别模型(如 large-v3)处理长音频,而系统可用资源不足。

解决方案:

  1. 使用更小的模型:在语音识别渠道中,将模型从 large-v3 更换为 mediumsmallbaselarge-v3 模型对显存和内存要求很高。
  2. 调整高级设置以降低资源占用:在菜单栏 工具/选项 -> 高级选项 中,进行如下修改:

    • CUDA数据类型: 将 float32 改为 float16int8
    • 识别准确度beam_size: 将 5 改为 1
    • 识别准确度best_of: 将 5 改为 1
    • 启用上下文感知: 将 true 改为 false
  3. 升级软件版本:您当前使用的是 v3.79 版本,建议升级到最新版 v3.98,新版本在资源管理和多显卡调度上有所优化。
  4. 检查并关闭其他占用显存的程序:确保没有其他大型应用(如游戏、设计软件)在后台运行,以释放显存。

请查阅相关文档:

Post Your Reply
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.

Related resource