madlad400-finetuned-loy-eng

This model is a fine-tuned version of jbochi/madlad400-3b-mt for translation from Lhowa to English.

Model details

• Developed by: SIL Global
• Finetuned from model: jbochi/madlad400-3b-mt
• Model type: Translation
• Source language: Lhowa (loy)
• Target language: English (eng)
• License: closed/private

Datasets

The model was trained on a parallel corpus of plain text files:

Lhowa:

• Lhowa Tibetan Bible
• License: © 2023, Wycliffe Bible Translators, Inc. and Nepal Bible Society. Used with permission.

English:

• English back-translation of Lhowa Bible
• License: © 2023, Wycliffe Bible Translators, Inc. and Nepal Bible Society. Used with permission.

Framework versions

• Transformers: nan

Usage

You can use this model with the transformers library like this:

from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

tokenizer = AutoTokenizer.from_pretrained("sil-ai/madlad400-finetuned-loy-eng")
model = AutoModelForSeq2SeqLM.from_pretrained("sil-ai/madlad400-finetuned-loy-eng")

inputs = tokenizer("Your input text here", return_tensors="pt")
outputs = model.generate(**inputs)
print(tokenizer.decode(outputs[0]))

madlad400-finetuned-loy-eng

This model is a fine-tuned version of jbochi/madlad400-3b-mt on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1027
• Chrf: 84.0833

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0005
• train_batch_size: 4
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 10.0

Training results

Training Loss	Epoch	Step	Validation Loss	Chrf
0.4467	4.8284	1600	0.3144	65.4472
0.1963	9.6567	3200	0.1041	83.7735

Framework versions

• PEFT 0.12.0
• Transformers 4.44.2
• Pytorch 2.4.1+cu124
• Datasets 2.21.0
• Tokenizers 0.19.1