text_shortening_model_v71

This model is a fine-tuned version of t5-small on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.3875
• Bert precision: 0.9031
• Bert recall: 0.9018
• Bert f1-score: 0.902
• Average word count: 6.4725
• Max word count: 18
• Min word count: 2
• Average token count: 10.5235
• % shortened texts with length > 12: 1.4014

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: 0.0003
• train_batch_size: 64
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 40

Training results

Training Loss	Epoch	Step	Validation Loss	Bert precision	Bert recall	Bert f1-score	Average word count	Max word count	Min word count	Average token count	% shortened texts with length > 12
1.8083	1.0	37	1.2558	0.8866	0.8836	0.8845	6.5606	17	1	10.4024	1.4014
1.3282	2.0	74	1.1453	0.8901	0.8914	0.8902	6.6797	18	1	10.5205	1.9019
1.1679	3.0	111	1.1190	0.8937	0.8928	0.8928	6.4825	18	1	10.4174	1.8018
1.0406	4.0	148	1.0827	0.8927	0.8956	0.8936	6.7377	18	2	10.6837	1.6016
0.9626	5.0	185	1.0821	0.8969	0.8978	0.8969	6.6176	18	2	10.6476	2.002
0.8814	6.0	222	1.0887	0.8974	0.9004	0.8984	6.7538	18	2	10.7908	2.3023
0.8163	7.0	259	1.0816	0.8972	0.8979	0.8971	6.6056	18	2	10.6096	1.8018
0.7636	8.0	296	1.0855	0.8987	0.8999	0.8988	6.5846	18	2	10.6967	2.002
0.7237	9.0	333	1.0949	0.8988	0.9004	0.8992	6.6346	18	2	10.6797	1.7017
0.6776	10.0	370	1.1174	0.9002	0.9017	0.9005	6.6186	18	2	10.6947	1.7017
0.6399	11.0	407	1.1237	0.8988	0.9002	0.8991	6.6316	18	2	10.6567	2.1021
0.5949	12.0	444	1.1426	0.8999	0.8988	0.8989	6.4755	18	2	10.5485	1.4014
0.5685	13.0	481	1.1564	0.9003	0.9015	0.9004	6.6216	18	2	10.6136	1.7017
0.5374	14.0	518	1.1690	0.9003	0.8997	0.8995	6.5506	18	2	10.5726	1.8018
0.5183	15.0	555	1.1736	0.9008	0.8997	0.8998	6.5415	18	2	10.5526	1.6016
0.4862	16.0	592	1.1882	0.8995	0.9001	0.8994	6.5936	18	2	10.6056	1.3013
0.4769	17.0	629	1.1910	0.9005	0.9003	0.8999	6.5716	18	2	10.6026	1.6016
0.4565	18.0	666	1.1957	0.9009	0.9	0.9	6.4615	18	2	10.5275	1.1011
0.4264	19.0	703	1.2276	0.9008	0.9004	0.9001	6.5125	18	2	10.5556	1.4014
0.4245	20.0	740	1.2415	0.9023	0.9005	0.9009	6.4605	18	2	10.4945	1.4014
0.4015	21.0	777	1.2658	0.9011	0.9004	0.9003	6.5135	18	2	10.5636	1.2012
0.3903	22.0	814	1.2779	0.9021	0.9018	0.9015	6.5495	18	2	10.5475	1.1011
0.3821	23.0	851	1.2899	0.9016	0.902	0.9014	6.5716	18	2	10.6336	1.4014
0.3595	24.0	888	1.3062	0.9007	0.9013	0.9005	6.5936	18	2	10.6947	1.3013
0.3551	25.0	925	1.3088	0.9015	0.9005	0.9006	6.4975	17	2	10.5355	1.2012
0.343	26.0	962	1.3169	0.9018	0.9009	0.9009	6.5005	17	2	10.5716	1.2012
0.3426	27.0	999	1.3264	0.8997	0.9018	0.9003	6.6486	17	2	10.7658	1.4014
0.3314	28.0	1036	1.3234	0.9018	0.9008	0.9008	6.4865	18	2	10.5165	1.2012
0.3187	29.0	1073	1.3378	0.9013	0.9003	0.9003	6.5055	18	2	10.5305	1.2012
0.3169	30.0	1110	1.3497	0.9015	0.9003	0.9004	6.4835	18	2	10.5546	1.2012
0.312	31.0	1147	1.3589	0.9018	0.8997	0.9003	6.4585	18	2	10.4615	1.2012
0.2995	32.0	1184	1.3572	0.901	0.9006	0.9004	6.5215	18	2	10.5866	1.3013
0.2987	33.0	1221	1.3647	0.9014	0.9009	0.9007	6.5305	18	2	10.5956	1.5015
0.2907	34.0	1258	1.3693	0.902	0.9007	0.9009	6.4585	18	2	10.5205	1.2012
0.2853	35.0	1295	1.3774	0.9026	0.9016	0.9016	6.4935	18	2	10.5385	1.2012
0.2746	36.0	1332	1.3815	0.9027	0.9023	0.9021	6.5285	18	2	10.5706	1.4014
0.2798	37.0	1369	1.3818	0.9026	0.9016	0.9016	6.4935	18	2	10.5285	1.4014
0.2801	38.0	1406	1.3858	0.9031	0.9018	0.902	6.4665	18	2	10.5175	1.3013
0.2773	39.0	1443	1.3868	0.9031	0.9018	0.902	6.4625	18	2	10.5185	1.3013
0.2756	40.0	1480	1.3875	0.9031	0.9018	0.902	6.4725	18	2	10.5235	1.4014

Framework versions

• Transformers 4.33.1
• Pytorch 2.0.1+cu118
• Datasets 2.14.5
• Tokenizers 0.13.3