Transcription fails when Audio files are too large
#2
by
JeanMi0LL
- opened
The demo functions quite good actually, thank you for this awesome work on this code!
However, it seems it may encounter issues when the audio input is too long (over 10 minutes).
I got the following error messages; but I cannot be sure whether they depend on my hardware, or if I could edit the code to comply with these constraints.
index-Bxw4oVts.js:2929 An error occurred during model execution: "Error: [WebGPU] Kernel "[Cast] /lm_head/Cast" failed. Error: Failed to generate kernel's output[0] with dims [1,8633,131072]. If you are running with pre-allocated output, please make sure the output type/dims are correct. Error: 266544008".
index-Bxw4oVts.js:2874 An uncaught WebGPU validation error was raised: Buffer size (4769836096) exceeds the max buffer size limit (2147483648).
- While calling [Device].CreateBuffer([BufferDescriptor]).
device.onuncapturederror @ index-Bxw4oVts.js:2874
Buffer size (4769836096) exceeds the max buffer size limit (2147483648).
- While calling [Device].CreateBuffer([BufferDescriptor]).