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

EN, 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: 0.6363
• Smatch Precision: 75.39
• Smatch Recall: 77.67
• Smatch Fscore: 76.51
• Smatch Unparsable: 0
• Percent Not Recoverable: 0.2129

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
0.3131	1.0	10431	1.5867	25.55	66.9	36.97	0	0.0194
0.0897	2.0	20862	1.0779	36.21	72.1	48.2	0	0.0968
0.1392	3.0	31294	0.7726	42.78	75.64	54.65	0	0.1936
0.085	4.0	41725	0.7040	46.38	76.85	57.85	0	0.0774
0.0008	5.0	52156	0.6874	47.47	76.12	58.47	0	0.1161
0.003	6.0	62588	0.6477	53.05	77.36	62.94	0	0.1742
0.0306	7.0	73019	0.6230	52.01	78.19	62.47	0	0.0968
0.0176	8.0	83451	0.6139	52.78	78.53	63.13	0	0.2129
0.0004	9.0	93882	0.6737	58.01	77.55	66.37	0	0.1355
0.0018	10.0	104313	0.6187	58.99	77.99	67.17	0	0.1161
0.0188	11.0	114745	0.6119	62.35	78.01	69.31	0	0.1161
0.0055	12.0	125176	0.6455	62.08	79.07	69.55	0	0.0774
0.0555	13.0	135607	0.6502	62.35	78.17	69.37	0	0.1355
0.0041	14.0	146039	0.6509	65.88	78.31	71.56	0	0.1742
0.0064	15.0	156470	0.6771	66.98	78.33	72.21	0	0.1355
0.0031	16.0	166902	0.6361	68.12	78.66	73.01	0	0.0774
0.0131	17.0	177333	0.6390	69.49	78.66	73.79	0	0.0968
0.0067	18.0	187764	0.6933	69.67	78.4	73.77	0	0.1549
0.0267	19.0	198196	0.6558	70.64	78.71	74.46	0	0.0774
0.0146	20.0	208627	0.6574	71.23	78.93	74.88	0	0.1161
0.0025	21.0	219058	0.6781	71.88	78.28	74.94	0	0.1936
0.0044	22.0	229490	0.6491	73.08	78.57	75.72	0	0.1161
0.0234	23.0	239921	0.6458	74.02	78.33	76.12	0	0.1549
0.0001	24.0	250353	0.6485	74.58	77.98	76.24	0	0.2129
0.0448	25.0	260775	0.6363	75.39	77.67	76.51	0	0.2129

Framework versions

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