nllb-lora-zh2Atayal

This model is a fine-tuned version of facebook/nllb-200-distilled-600M on the None dataset. It achieves the following results on the evaluation set:

• Loss: 7.1847

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.001
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
7.0315	1.0	240	7.3253
6.9621	2.0	480	7.2832
6.9505	3.0	720	7.2591
6.9091	4.0	960	7.2445
6.8988	5.0	1200	7.2312
6.9053	6.0	1440	7.2217
6.8806	7.0	1680	7.2163
6.8691	8.0	1920	7.2096
6.8917	9.0	2160	7.2044
6.8568	10.0	2400	7.1998
6.8459	11.0	2640	7.1964
6.8482	12.0	2880	7.1950
6.867	13.0	3120	7.1915
6.8377	14.0	3360	7.1902
6.8557	15.0	3600	7.1890
6.8409	16.0	3840	7.1872
6.8447	17.0	4080	7.1870
6.8176	18.0	4320	7.1849
6.8346	19.0	4560	7.1852
7.1757	19.9186	4780	7.1847

Framework versions

• PEFT 0.15.0
• Transformers 4.51.2
• Pytorch 2.2.2+cu118
• Datasets 3.5.0
• Tokenizers 0.21.1