myduy/vinai-vinai-translate-vi2en-v2-ckpt40584-061201

vi_mbart_en2amr

This model is a fine-tuned version of facebook/mbart-large-50-many-to-many-mmt on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.3850
• Smatch Precision: 79.44
• Smatch Recall: 82.04
• Smatch Fscore: 80.72
• Smatch Unparsable: 49
• Percent Not Recoverable: 0.1248

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: 8
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 25

Training results

Training Loss	Epoch	Step	Validation Loss	Percent Not Recoverable	Smatch Fscore	Smatch Precision	Smatch Recall	Smatch Unparsable
0.2877	0.9995	1623	3.1858	0.0	8.07	4.39	50.65	1125
0.2046	1.9995	3246	2.4277	0.0	2.03	1.05	27.91	1560
0.144	2.9995	4869	2.0027	0.0	4.65	2.48	38.63	958
0.0958	3.9995	6492	1.4330	0.0	4.72	2.51	39.29	812
0.0833	4.9995	8115	1.1425	0.3743	13.28	7.41	63.63	107
0.0701	5.9995	9738	1.0069	0.2495	11.22	6.16	62.84	153
0.0545	6.9995	11361	0.9798	0.0	26.35	16.0	74.6	76
0.037	7.9995	12984	0.8805	0.9981	17.69	10.16	68.02	65
0.0328	8.9995	14607	0.6323	2.2458	30.35	19.02	74.99	33
0.0315	9.9995	16230	0.6267	5.4273	24.37	14.64	72.82	85
0.0315	9.9995	16239	0.6085	16.38	73.69	26.8	80	7.7979
0.0173	10.9995	17862	0.5898	31.48	78.39	44.92	72	6.0512
0.0155	11.9995	19485	0.5628	43.89	82.41	57.28	77	0.3119
0.0121	12.9995	21108	0.4881	52.89	82.32	64.4	92	0.1248
0.0131	13.9995	22731	0.4575	53.13	83.25	64.87	72	0.0
0.0136	14.9995	24354	0.4365	66.87	83.99	74.46	73	0.1248
0.0096	15.9995	25977	0.4532	69.72	82.86	75.72	65	0.0624
0.0144	16.9995	27600	0.4128	72.61	83.11	77.5	77	0.0624
0.0026	17.9995	29223	0.4037	74.81	82.73	78.57	63	0.1871
0.0027	18.9995	30846	0.3926	77.05	82.38	79.62	53	0.2495
0.0034	19.9995	32469	0.4063	77.24	82.05	79.57	53	0.1248
0.0013	20.9995	34092	0.3955	78.7	82.89	80.74	50	0.1871
0.0018	21.9995	35715	0.3949	78.73	82.47	80.55	49	0.1871
0.0013	22.9995	37338	0.3920	78.76	82.27	80.47	52	0.1248
0.0002	23.9995	38961	0.3856	79.55	82.23	80.87	50	0.1248
0.0003	24.9995	40584	0.3850	79.44	82.04	80.72	49	0.1248

Framework versions

• Transformers 4.57.2
• Pytorch 2.9.1+cu128
• Datasets 4.4.1
• Tokenizers 0.22.1