Modelpublic

KnoxDevelopers/English-to-Kikuyu-language-translation-LoRA

sourceHugging Facecc-by-nc-4.0updated 3mo agoView on Hugging Face

0likes13downloads

Model Card

English-to-Kikuyu-language-translation-LoRA

This model is a fine-tuned version of facebook/nllb-200-distilled-600M on the English-Kikuyu dataset. It achieves the following results on the evaluation set:

—Loss: 1.8012

Model description

A LoRA adapter for translating text from English to Kikuyu.

Intended uses & limitations

—Translate sentences, short and medium long paragraphs.
—Limited to translation tasks only.
—Limited output length (max length = 512)

Training and evaluation data

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

—learning_rate: 0.0002
—trainbatchsize: 8
—evalbatchsize: 8
—seed: 42
—optimizer: Use OptimizerNames.ADAMWTORCHFUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
—lrschedulertype: linear
—num_epochs: 5
—mixedprecisiontraining: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
2.1453	1.0	3383	1.9200
1.9903	2.0	6766	1.8605
2.0602	3.0	10149	1.8271
2.0325	4.0	13532	1.8073
1.9727	5.0	16915	1.8012

Evaluation metrics

Metric	Results	Explained
sacrebleu	12.974	Eval matches exact word sequence referencing our ground truth and adapter output.
		0-20= translation is literal or broken, 20-50=good (sweet spot), >60=very good or identical training and test data.
chrF++	41.023	Eval matches character sequence instead of whole words.
		Model adapter is getting the root words and grammar concepts right but using different word variations or spellings than our ground truth, this is concluded due to the low sacrebleu results.

Framework versions

—PEFT 0.19.1
—Transformers 5.12.1
—Pytorch 2.11.0+cu128
—Datasets 4.0.0
—Tokenizers 0.22.2