BramVanroy/mbart-large-cc25-ft-amr30-es_nl

ES and NL to AMR parsing

This model is a fine-tuned version of facebook/mbart-large-cc25 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.6542
• Smatch Precision: 73.41
• Smatch Recall: 76.04
• Smatch Fscore: 74.7
• Smatch Unparsable: 0
• Percent Not Recoverable: 0.2613

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: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 25

Training results

Training Loss	Epoch	Step	Validation Loss	Smatch Precision	Smatch Recall	Smatch Fscore	Smatch Unparsable	Percent Not Recoverable
0.2675	1.0	6954	1.3790	23.26	65.74	34.36	0	0.0
0.1137	2.0	13908	1.0480	32.79	71.81	45.02	0	0.0
0.1606	3.0	20862	0.8573	38.99	72.53	50.72	0	0.0581
0.0923	4.0	27817	0.7614	40.4	75.22	52.56	0	0.0290
0.0292	5.0	34771	0.7935	46.44	75.63	57.54	0	0.0290
0.0106	6.0	41725	0.7326	49.54	75.8	59.92	0	0.0
0.0054	7.0	48679	0.6385	51.35	76.11	61.33	0	0.0290
0.048	8.0	55634	0.6489	53.03	76.79	62.74	0	0.0581
0.0334	9.0	62588	0.6128	59.05	77.3	66.95	0	0.0581
0.0393	10.0	69542	0.6242	57.91	77.02	66.11	0	0.0871
0.0251	11.0	76496	0.6417	58.46	77.31	66.58	0	0.1742
0.0035	12.0	83451	0.6271	62.28	76.99	68.86	0	0.0581
0.0228	13.0	90405	0.6685	62.47	76.97	68.97	0	0.1452
0.0119	14.0	97359	0.6414	63.12	77.23	69.47	0	0.1161
0.0066	15.0	104313	0.6515	65.63	77.02	70.87	0	0.0871
0.0025	16.0	111268	0.6467	67.05	77.35	71.83	0	0.0871
0.0024	17.0	118222	0.6657	65.47	77.13	70.82	0	0.0581
0.0223	18.0	125176	0.6754	67.56	77.21	72.06	0	0.1452
0.034	19.0	132130	0.6569	68.47	76.97	72.47	0	0.1161
0.007	20.0	139085	0.6734	69.86	77.17	73.34	0	0.2033
0.0224	21.0	146039	0.6544	70.95	76.72	73.72	0	0.1742
0.005	22.0	152993	0.6619	72.18	76.83	74.43	0	0.1742
0.0055	23.0	159947	0.6683	72.21	76.42	74.26	0	0.2323
0.0	24.0	166902	0.6585	72.8	76.3	74.51	0	0.2033
0.0693	25.0	173850	0.6542	73.41	76.04	74.7	0	0.2613

Framework versions

• Transformers 4.34.0.dev0
• Pytorch 2.0.1+cu117
• Datasets 2.14.2
• Tokenizers 0.13.3