Is MPS slow for transcription? MacBook Pro M2 16 GB but model.safetensors rate is 3.48MB/s

[jasonrichdarmawan]

My goal is to translate a video from Chinese to English.

What I did:

1. Install decipher
2. run decipher gui

$ decipher gui        
Running on local URL:  http://127.0.0.1:7860

To create a public link, set `share=True` in `launch()`.
Extracting audio from video: 100%|████████████████████████████████████████████████████████████████| 1378/1378 [00:12<00:00, 108.63it/s]
MPS is being used for this transcription, this process may take a while.
model.safetensors:  42%|███████████████████████████████▏                                          | 1.30G/3.09G [07:23<16:21, 1.82MB/s]

Expected result: lower than Google Colab with NVIDIA T4 which is 214 MB/s

model.safetensors: 100% 3.09G/3.09G [00:14<00:00, 214MB/s]

Actual result: 3.48 MB/s

MPS is being used for this transcription, this process may take a while.
model.safetensors:  42%|███████████████████████████████▏                                          | 1.30G/3.09G [07:23<16:21, 1.82MB/s]

[phineas-pta]

it's your internet speed

[jasonrichdarmawan]

This is run locally, are you sure that it's my internet speed?

[phineas-pta]

because it's downloading the checkpoint