Automatic Speech Recognition
Transformers
PyTorch
JAX
Safetensors
Korean
hubert
feature-extraction
speech
audio
custom_code
Instructions to use team-lucid/hubert-base-korean with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use team-lucid/hubert-base-korean with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("automatic-speech-recognition", model="team-lucid/hubert-base-korean", trust_remote_code=True)# Load model directly from transformers import AutoProcessor, AutoModel processor = AutoProcessor.from_pretrained("team-lucid/hubert-base-korean", trust_remote_code=True) model = AutoModel.from_pretrained("team-lucid/hubert-base-korean", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
training time
#4
by beginnerAI - opened
hello,
i want to train this model for other languages.
may i ask the training time and your computer? i would like to know how long does it take if i train it in NVIDIA A100 or 4090 etc.
thanks in advance.
seb
Thanks for your interest.
We used TPU v3-8 to train our model and it took about 40 and 84 hours for phase 1 and phase 2 of the base model, respectively.
I don't know how long it will take in other cases because the environment I used is not common called JAX.
If necessary, I will share the code used to train the model through GitHub.
hyunwoo3235 changed discussion status to closed