Repository Issue Activity (beta)

#279 Feature: Add FunASR/SenseVoice as a non-autoregressive backend

#186 How about some honest speed/quality tests for a change

#282 Benchmark: FunASR SenseVoice — 5-10x faster than Whisper with comparable accuracy

#281 Benchmark comparison: SenseVoice (non-autoregressive) vs Whisper

#280 Feature Request: Benchmark SenseVoice — 170x realtime non-autoregressive ASR

#274 Benchmark shows speed, but what about words error rate?

#276 Specs requirements

#255 Mac M1 - multiple issues preventing transcription

#247 Prediction failed: attempt to get argmin of an empty sequence

#271 Dual Licensing Confusion

#270 Error installing via pipx on macOS Intel

#268 Running in an offline format

#266 With Flash Attention: shape '[3, 1, 20, 64]' is invalid for input of size 5760000

#264 Deleted

#263 Support for max_line_width and max_line_count

#121 [Question]How to translate to another specified language

#262 why this issue with numpy

#253 You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`.

#257 The kernel is getting restarted when I run faster whisper large-v2 model

#245 flash_attention_2 on macOS?

#244 not working on macOS

#36 Support for CPU Mode

#168 Is there an api to get the transcription progress？

#256 Would be nice a CPU only version too.

#103 I don't have a gpu. Is there an option to run only on CPU?