MarianMT-Nepali-to-English-Synthetic-Pretrain-Continued

This model is a fine-tuned version of iamTangsang/MarianMT-Nepali-to-English on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.0186
• Bleu: 30.3843
• Gen Len: 73.8617

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: 5e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Bleu	Gen Len	Validation Loss
1.1924	0.0561	12000	26.1232	74.765	1.3707
1.0978	0.1122	24000	26.0469	78.0433	1.2622
1.0316	0.1684	36000	28.0954	73.6692	1.2075
0.9993	0.2245	48000	28.1497	75.1633	1.1745
0.9651	0.2806	60000	29.1482	75.9908	1.1428
0.9415	0.3367	72000	27.0537	82.6383	1.1183
0.9251	0.3928	84000	27.637	79.2592	1.0864
0.9008	0.4489	96000	29.0405	76.6583	1.0683
0.8726	0.5051	108000	29.923	75.4483	1.0494
0.8701	0.5612	120000	29.2328	77.2858	1.0316
0.8546	0.6173	132000	29.6585	76.1308	1.0185
0.8392	0.6734	144000	30.5079	78.0417	1.0072
0.9316	0.7295	156000	29.264	71.7958	1.1055
0.9008	0.7857	168000	27.2107	80.45	1.0779
0.8982	0.8418	180000	27.7493	78.6583	1.0645
0.8655	0.8979	192000	28.2546	79.6033	1.0438
0.8654	0.9540	204000	1.0324	30.2355	73.8317

Framework versions

• Transformers 4.45.2
• Pytorch 2.5.1+cu121
• Datasets 3.2.0
• Tokenizers 0.20.3