mt5-small-finetuned-arith

This model is a fine-tuned version of google/mt5-small on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.6651
• Rouge1: 90.0
• Rouge2: 70.4082
• Rougel: 85.3061
• Rougelsum: 85.102

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: 5.6e-05
• train_batch_size: 128
• eval_batch_size: 128
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 64

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum
No log	1.0	7	11.7623	0.0	0.0	0.0	0.0
No log	2.0	14	11.0473	0.2041	0.0	0.2041	0.2041
No log	3.0	21	9.4965	0.4082	0.0	0.4082	0.4082
No log	4.0	28	8.3848	0.8673	0.0	0.8673	0.8673
No log	5.0	35	7.6170	1.7515	0.0	1.7114	1.6753
No log	6.0	42	7.0008	4.9101	0.0	4.9093	4.8585
No log	7.0	49	6.7836	8.0777	0.0	7.7956	7.9186
16.7453	8.0	56	6.6780	12.3572	0.0	12.1332	11.878
16.7453	9.0	63	5.2800	13.5863	0.1701	12.7907	12.8991
16.7453	10.0	70	4.4990	13.8751	0.1701	13.1962	13.1834
16.7453	11.0	77	4.3624	13.4276	0.1701	13.3009	13.2722
16.7453	12.0	84	4.1101	14.0537	0.3401	13.3534	13.354
16.7453	13.0	91	3.7171	14.2128	0.3401	13.4985	13.4888
16.7453	14.0	98	3.4322	13.9164	0.1701	13.3916	13.3625
16.7453	15.0	105	3.2408	13.931	0.3401	13.7998	13.7901
6.4188	16.0	112	3.0734	14.0816	0.3401	13.7901	13.7901
6.4188	17.0	119	2.9270	14.344	0.8242	14.1983	14.208
6.4188	18.0	126	2.7746	16.7178	2.4928	16.3946	16.4334
6.4188	19.0	133	2.6117	22.7164	7.4678	22.1643	22.1381
6.4188	20.0	140	2.4419	25.0641	9.4306	24.2861	24.2714
6.4188	21.0	147	2.2793	32.0373	13.6803	31.0317	30.8515
6.4188	22.0	154	2.0741	40.1666	21.0894	38.5458	38.4592
6.4188	23.0	161	1.8635	40.1133	21.1222	38.1971	38.1165
3.1581	24.0	168	1.6788	47.1732	25.3843	44.6854	44.6021
3.1581	25.0	175	1.5153	49.4894	27.0538	46.9745	46.8775
3.1581	26.0	182	1.3337	47.7463	25.9589	45.3779	45.2896
3.1581	27.0	189	1.1634	48.6608	26.067	46.293	46.1794
3.1581	28.0	196	1.0392	86.6181	65.5782	81.9242	81.8732
3.1581	29.0	203	0.9519	90.0	70.4082	85.3061	85.102
3.1581	30.0	210	0.8837	90.0	70.4082	85.3061	85.102
3.1581	31.0	217	0.8246	90.0	70.4082	85.3061	85.102
2.0354	32.0	224	0.7630	90.0	70.4082	85.3061	85.102
2.0354	33.0	231	0.7221	90.0	70.4082	85.3061	85.102
2.0354	34.0	238	0.6957	90.0	70.4082	85.3061	85.102
2.0354	35.0	245	0.6852	90.0	70.4082	85.3061	85.102
2.0354	36.0	252	0.6734	90.0	70.4082	85.3061	85.102
2.0354	37.0	259	0.6667	90.0	70.4082	85.3061	85.102
2.0354	38.0	266	0.6670	90.0	70.4082	85.3061	85.102
2.0354	39.0	273	0.6684	90.0	70.4082	85.3061	85.102
1.5363	40.0	280	0.6626	90.0	70.4082	85.3061	85.102
1.5363	41.0	287	0.6621	90.0	70.4082	85.3061	85.102
1.5363	42.0	294	0.6699	90.0	70.4082	85.3061	85.102
1.5363	43.0	301	0.6751	90.0	70.4082	85.3061	85.102
1.5363	44.0	308	0.6839	90.0	70.4082	85.3061	85.102
1.5363	45.0	315	0.6987	90.0	70.4082	85.3061	85.102
1.5363	46.0	322	0.7060	90.0	70.4082	85.3061	85.102
1.5363	47.0	329	0.7125	90.0	70.4082	85.3061	85.102
1.324	48.0	336	0.7103	90.0	70.4082	85.3061	85.102
1.324	49.0	343	0.7098	90.0	70.4082	85.3061	85.102
1.324	50.0	350	0.7088	90.0	70.4082	85.3061	85.102
1.324	51.0	357	0.7112	90.0	70.4082	85.3061	85.102
1.324	52.0	364	0.7094	90.0	70.4082	85.3061	85.102
1.324	53.0	371	0.7041	90.0	70.4082	85.3061	85.102
1.324	54.0	378	0.6939	90.0	70.4082	85.3061	85.102
1.2374	55.0	385	0.6843	90.0	70.4082	85.3061	85.102
1.2374	56.0	392	0.6791	90.0	70.4082	85.3061	85.102
1.2374	57.0	399	0.6755	90.0	70.4082	85.3061	85.102
1.2374	58.0	406	0.6715	90.0	70.4082	85.3061	85.102
1.2374	59.0	413	0.6661	90.0	70.4082	85.3061	85.102
1.2374	60.0	420	0.6639	90.0	70.4082	85.3061	85.102
1.2374	61.0	427	0.6629	90.0	70.4082	85.3061	85.102
1.2374	62.0	434	0.6635	90.0	70.4082	85.3061	85.102
1.199	63.0	441	0.6646	90.0	70.4082	85.3061	85.102
1.199	64.0	448	0.6651	90.0	70.4082	85.3061	85.102

Framework versions

• Transformers 4.33.1
• Pytorch 1.12.1
• Datasets 2.14.5
• Tokenizers 0.13.3