Downloads · 30 days
1.3K
4% of all-time downloads
marianmt-th-zh_cn
Lalita/marianmt-th-zh_cn
marianmt-th-zh_cn is a translation model from Lalita. Use it when you need text moved from one language to another. It is set up for transformers.
source languages: th target languages: zhcn dataset: model: transformer-align pre-processing: normalization + SentencePiece test set scores: 15.53
All-time downloads
28.3K
Public
Repo size
485 MB
Likes
1
Public
Repo makeup
Click a slice to open those files.
.bin242 MB · 99%
At a glance
- Task
- Translation
- Library
- transformers
- Model type
- marian
- Access
- Public
- Created
- Mar 2, 2022
- Updated
- Jun 29, 2021
- SHA
- bab48ced
Try a prompt
Model card
From the Hugging Face model README
marianmt-th-zh_cn
- source languages: th
- target languages: zh_cn
- dataset:
- model: transformer-align
- pre-processing: normalization + SentencePiece
- test set scores: 15.53
Training
Training scripts from LalitaDeelert/NLP-ZH_TH-Project. Experiments tracked at cstorm125/marianmt-th-zh_cn.
export WANDB_PROJECT=marianmt-th-zh_cn
python train_model.py --input_fname ../data/v1/Train.csv \\\\\\\\
\\\\t--output_dir ../models/marianmt-th-zh_cn \\\\\\\\
\\\\t--source_lang th --target_lang zh \\\\\\\\
\\\\t--metric_tokenize zh --fp16
Usage
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
tokenizer = AutoTokenizer.from_pretrained("Lalita/marianmt-zh_cn-th")
model = AutoModelForSeq2SeqLM.from_pretrained("Lalita/marianmt-zh_cn-th").cpu()
src_text = [
'ฉันรักคุณ',
'ฉันอยากกินข้าว',
]
translated = model.generate(**tokenizer(src_text, return_tensors="pt", padding=True))
print([tokenizer.decode(t, skip_special_tokens=True) for t in translated])
> ['我爱你', '我想吃饭。']
Requirements
transformers==4.6.0
torch==1.8.0
- Task
- Translation
- Library
- transformers
- Type
- marian
- Created
- Mar 2, 2022
- Updated
- Jun 29, 2021