Downloads · 30 days
37
10% of all-time downloads
ofu-ai/mbart-large-50-mmt-ko-vi
mbart-large-50-mmt-ko-vi is a translation model from ofu-ai. Use it when you need text moved from one language to another. It is set up for transformers. The card lists the license as mit.
This model is fine-tuned from mBART-large-50 using multilingual translation data of Korean legal documents for Korean-to-Vietnamese translation tasks.
Downloads · 30 days
37
10% of all-time downloads
All-time downloads
371
Public
Parameters
611M
2.5 GB on disk
Likes
3
Public
Click a slice to open those files.
.safetensors2.4 GB · 99%
From the Hugging Face model README
This model is fine-tuned from mBART-large-50 using multilingual translation data of Korean legal documents for Korean-to-Vietnamese translation tasks.
This model is used for text translation from Korean to Vietnamese.
This model is not suitable for translation tasks involving languages other than Korean.
The model may contain biases inherited from the training data and may produce inappropriate translations for sensitive topics.
The model was trained using multilingual translation data of Korean legal documents provided by AI Hub.
The following hyperparameters were used during training:
0.00018 (per device)8 (per device)42single-node (since _n_gpu=1 and no distributed training setup is indicated)1 (single NVIDIA GPU: RTX 4090)432 (calculated as train_batch_size * gradient_accumulation_steps)8 (evaluation does not use gradient accumulation)AdamW (indicated by optim=OptimizerNames.ADAMW_TORCH)linear (indicated by lr_scheduler_type=SchedulerType.LINEAR)1003The evaluation used a dataset partially extracted from Korean labor law precedents.
Model Architecture: Based on mBART-large-50, a multilingual sequence-to-sequence transformer model designed for translation tasks. The architecture includes 24 encoder and 24 decoder layers with 1,024 hidden units.
Software:
Hardware: NVIDIA RTX 4090 with 24GB VRAM was used for training and inference.
Tokenization and Preprocessing: The tokenization was performed using the SentencePiece model pre-trained with mBART-large-50. Text preprocessing included removing special characters, unnecessary whitespace, and normalizing line breaks.
Currently, there are no papers or blog posts available for this model.