test_llm_nllb_colab_100_b2_e_12clrlinearreload

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: 0.5442
• Rouge1: 0.6187
• Rouge2: 0.3907
• Rougel: 0.573
• Sacrebleu: 23.8256

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: 3e-05
• train_batch_size: 2
• eval_batch_size: 2
• seed: 237
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 12
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Sacrebleu
0.5035	1.0	2040	0.4985	0.5886	0.3527	0.5463	21.1436
0.4211	2.0	4080	0.4785	0.6043	0.3695	0.5589	22.5082
0.3277	3.0	6120	0.4776	0.6153	0.3792	0.5695	22.6802
0.2883	4.0	8160	0.4826	0.6142	0.3823	0.5695	23.3872
0.2641	5.0	10200	0.4900	0.6215	0.3881	0.5752	23.6105
0.2295	6.0	12240	0.5038	0.6166	0.3844	0.5709	23.3196
0.185	7.0	14280	0.5126	0.6155	0.3839	0.5704	23.4375
0.1777	8.0	16320	0.5230	0.6176	0.3867	0.5722	23.902
0.1445	9.0	18360	0.5305	0.621	0.3895	0.5735	23.7867
0.1354	10.0	20400	0.5373	0.6138	0.3825	0.5681	23.5844
0.1167	11.0	22440	0.5407	0.6173	0.3886	0.573	23.8997
0.1115	12.0	24480	0.5442	0.6187	0.3907	0.573	23.8256

Framework versions

• Transformers 4.41.2
• Pytorch 2.3.0+cu121
• Datasets 2.21.0
• Tokenizers 0.19.1