search

linto-ai / whisper-timestamped

Multilingual Automatic Speech Recognition with word-level timestamps and confidence
GNU Affero General Public License v3.0
1.87k stars 149 forks source link

How to use cuda? #174

Closed eliich closed 6 months ago

Jeronymous commented 6 months ago

The device is chosen when loading the model.

import whisper_timestamped as whisper
model = whisper.load_model("tiny", device="cuda")

and there is an option --device in the CLI

(same syntax as pytorch supported: "cuda:0", ...)

eliich commented 6 months ago

The device is chosen when loading the model.

import whisper_timestamped as whisper
model = whisper.load_model("tiny", device="cuda")

and there is an option --device in the CLI

(same syntax as pytorch supported: "cuda:0", ...)

It seems that when I do that cuda is not available. which is weird considering I used whisper and whisperX with cuda just fine