-
안녕하세요,
올려주신 블로그 포스팅이 많은 도움이 되어 저도 무사히 훈련을 시작했습니다, 감사합니다.
저는 ksponspeech 데이터를 다운 받아서 3080ti로 직접 training을 하고 있는데,
혹시 30만개 데이터셋 기준으로 epoch을 몇 번까지 돌리신 후에 해당 결과를 얻으셨는지 궁금해서 여쭙니다.
감사합니다 : )
-
https://github.com/espnet/espnet/blob/3651c2e67126c4544820cf148407be7f2679866c/espnet/nets/pytorch_backend/transformer/layer_norm.py#L12-L24
I discovered that the epsilon of LayerNorm in espnet set…
-
Hello,
The paper of CitriNet said they used 32 V100 GPUs, with a batch size of 32 per GPU.
then, the gradients of each GPU are accumulated and updated together?
When I trained CitriNet-384 mod…
-
### Describe the bug
The LM template forgets to call `hparams["pretrainer"].load_collected`, the way that recipes such as `recipes/KsponSpeech/LM/train.py` do. This results in empty training data.
#…
-
hi, I am having trouble preparing data for ksponspeech. (https://github.com/speechbrain/speechbrain/tree/89553c919393ebf45046a47493a15be359cbd009/recipes/KsponSpeech)
I can't run ksponspeech_prepa…
-
### Describe the bug
There are some typos in below sources,
recipes/LibriSpeech/Tokenizer/train.py
recipes/KsponSpeech/Tokenizer/train.py
.
In comment:
"""
~
coverts -> converts
bas -> bad…
-
## Title
- 패키지 설치 오류
안녕하세요.
torchaudio: pip install torchaudio==0.6.0 (Refer [here](https://github.com/pytorch/pytorch) for problem installing torchaudio)
를 실행하였을 때 0.6.0 버전이 존재하지 않아 가장 낮은 버전인…
-
Ksponspeech 관련 질문입니다.
코드 관련 질문이 아닌 점 죄송합니다.
학습 이후, inference.py 에서 pcm파일을 여는 과정 중 발생한 문제입니다.
```
def load_audio(audio_path: str, del_silence: bool = False, extension: str = 'pcm') -> np.ndar…
-
### Describe the bug
When I run fine tuning ASR training with mulit-GPU(DDP), below error occures.
FileExistsError: [Errno 17] File exists: '/workspace/speechbrain/recipes/KdialectSpeech/ASR/Confo…
-
제 윈도우 데스크탑으로 훈련시키면 오류없이 잘 됩니다 하지만 구글 CoLab을 사용하여 데이터셋을 구글드라이브에 올려놓고 아래 명령어로 훈련시켰더니
python ./bin/main.py model=ds2 train=ds2_train train.dataset_path="/content/drive/MyDrive/kspon"
[2022-06-24 05…