pyVideoTrans

English

简体中文

English

简体中文

Appearance

Understanding and Fixing the 'Cannot find an appropriate cached snapshot folder for the specified revision...' Error ​

When running the pyVideoTrans software in faster-whisper mode, you might frequently encounter the complex-looking error described below. Please don't worry.

This is typically not a software bug, but rather a common network issue.

Traceback (most recent call last):
  File "videotrans\task\_only_one.py", line 43, in run
  File "videotrans\task\trans_create.py", line 371, in recogn
  File "videotrans\recognition\__init__.py", line 236, in run
  File "videotrans\recognition\_base.py", line 75, in run
  File "videotrans\recognition\_overall.py", line 155, in _exec
RuntimeError: err[msg]=Cannot find an appropriate cached snapshot folder for the specified revision on the local disk and outgoing traffic has been disabled. To enable repo look-ups and downloads online, pass 'local_files_only=False' as input.

I. What Exactly Does This Error Mean? ​

Simply put, this error is like a two-step chain reaction:

  1. Step One: Model Download Failure. The program needs to download the "faster-whisper" model from the internet to function. However, due to your network's inability to connect smoothly to the model server (Hugging Face), or a download interruption, the model failed to download successfully.

  2. Step Two: Model Not Found in Offline Mode. When the faster-whisper library detects a network download failure, it smartly attempts to locate the model in your computer's local folders (this is called "offline mode"). But since the initial download in Step One failed, your computer's local folders are naturally empty, leading to the "model not found locally" report and ultimately the error you see.

In a nutshell: A poor network connection leads to model download failure, and download failure results in no models being available locally.

II. How to Resolve This Issue? ​

The core idea is: Ensure the model is completely downloaded to your computer. Here are two simple and effective solutions.

The program has built-in channels for downloading from domestic mirror sites, which is usually the fastest and most stable method.

  1. Disconnect Proxy/VPN: If you are using any form of system proxy or VPN, please turn them off first. VPNs can be unstable and might disconnect during long downloads of large models, leading to failure.

  2. Delete Failed Cache Files:

    • Locate the program's main folder (i.e., the folder where sp.exe is located).
    • Navigate into the models subfolder.
    • Find and delete any folders related to the model you attempted to download (e.g., models--Systran--faster-whisper-largexxx). This step clears incomplete or corrupted download files.

  1. Rerun the Program: After disabling the proxy and clearing the cache, restart the main program. It should now be able to download the model smoothly from the domestic mirror server.

Solution Two: Manually Download Models Using a Dedicated Tool ​

If Solution One doesn't work, or if your network environment is unusual, we've provided a specialized model download tool. You can download all the necessary models (8 faster-whisper models in total) in one go, just like downloading regular software, before running the main program.

  1. Download the Tool:

    • For Chinese Users (Recommended):

    • For Overseas Users or Those with Stable Proxy:

      • Click to download down_huggingface.exe
      • (Note: This version downloads directly from the Hugging Face official website, requiring stable access to the international internet or a properly configured system proxy.)

  2. Usage Instructions:

    • Place the downloaded .exe file into the main program's folder (next to sp.exe).
    • Double-click to run the .exe file you downloaded. It will open a command-line window and automatically begin downloading all necessary models.
    • The download process may take some time, especially for the large series models which are very big files. Please wait patiently until the window indicates completion.

Important Note:

  • Models like large-v1, large-v2, large-v3, etc., are very large (several GBs). If your network is unstable, the download might fail midway.
  • If the download fails, don't worry. Simply double-click to run the same .exe file again. It supports resume functionality and will continue downloading from where it left off.

Once all models have been successfully downloaded to the models folder using either of the methods above, rerun the main program, and the issue should be resolved!