text_shortening_model_v77

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

• Loss: 1.4183
• Bert precision: 0.8989
• Bert recall: 0.9008
• Bert f1-score: 0.8994
• Average word count: 6.9571
• Max word count: 15
• Min word count: 2
• Average token count: 11.2896
• % shortened texts with length > 12: 2.0859

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
1.6557	1.0	30	1.2281	0.8905	0.8844	0.887	6.5926	15	1	10.4699	1.4724
1.2524	2.0	60	1.1200	0.8933	0.8902	0.8913	6.7227	15	2	10.8294	0.9816
1.1024	3.0	90	1.0914	0.8952	0.8931	0.8937	6.7706	16	2	10.811	1.1043
0.9921	4.0	120	1.0805	0.8935	0.8943	0.8935	6.9141	17	2	11.0454	1.3497
0.8865	5.0	150	1.1025	0.8971	0.8949	0.8956	6.7607	16	2	10.9828	1.4724
0.8273	6.0	180	1.1039	0.9005	0.8989	0.8993	6.7583	16	2	10.8982	1.9632
0.7585	7.0	210	1.0975	0.8981	0.8991	0.8982	6.9497	16	3	11.135	2.2086
0.7086	8.0	240	1.1068	0.8989	0.8971	0.8976	6.7374	15	2	10.9411	1.1043
0.6541	9.0	270	1.1340	0.898	0.902	0.8996	7.1239	17	2	11.4258	2.8221
0.6167	10.0	300	1.1316	0.8981	0.8996	0.8984	6.962	16	2	11.1436	2.6994
0.5817	11.0	330	1.1507	0.8984	0.8995	0.8985	6.9264	16	2	11.2466	2.3313
0.547	12.0	360	1.1416	0.899	0.8993	0.8988	6.8601	15	2	11.1865	1.5951
0.5181	13.0	390	1.1775	0.8987	0.8993	0.8986	6.8969	15	3	11.1571	1.8405
0.4874	14.0	420	1.2131	0.8969	0.8999	0.8979	7.0638	16	3	11.3877	2.454
0.461	15.0	450	1.2263	0.9009	0.9001	0.9001	6.8871	15	2	11.1497	1.1043
0.4449	16.0	480	1.2215	0.898	0.8998	0.8984	7.0184	15	2	11.3509	2.2086
0.4148	17.0	510	1.2528	0.8991	0.8997	0.899	6.8748	15	2	11.1939	1.227
0.3967	18.0	540	1.2512	0.8989	0.9006	0.8994	6.9853	15	2	11.3227	1.3497
0.385	19.0	570	1.2683	0.898	0.8992	0.8982	6.9755	17	2	11.3227	2.454
0.3667	20.0	600	1.2804	0.898	0.8982	0.8976	6.8528	15	2	11.1337	1.7178
0.3579	21.0	630	1.2892	0.9006	0.899	0.8993	6.762	15	2	11.0761	1.8405
0.3358	22.0	660	1.3081	0.8983	0.8996	0.8985	6.8957	15	2	11.3166	1.5951
0.3261	23.0	690	1.3189	0.8988	0.8997	0.8988	6.9571	15	2	11.2491	2.6994
0.3222	24.0	720	1.3116	0.8975	0.9	0.8983	7.027	15	2	11.3926	2.9448
0.2977	25.0	750	1.3295	0.8988	0.9005	0.8992	6.9755	15	2	11.308	1.9632
0.2977	26.0	780	1.3368	0.8972	0.9009	0.8986	7.0442	15	2	11.4184	2.5767
0.2795	27.0	810	1.3515	0.9015	0.9025	0.9015	6.9117	15	2	11.2601	2.2086
0.2758	28.0	840	1.3645	0.9018	0.9013	0.9011	6.8466	15	2	11.1264	1.9632
0.2696	29.0	870	1.3675	0.899	0.9002	0.8991	6.9399	15	2	11.2687	1.9632
0.261	30.0	900	1.3896	0.8986	0.901	0.8993	7.0282	17	2	11.3902	2.0859
0.2632	31.0	930	1.3802	0.8971	0.8988	0.8975	6.9546	15	2	11.2528	2.3313
0.2488	32.0	960	1.3919	0.8998	0.9011	0.9	6.9656	15	2	11.2748	2.454
0.2467	33.0	990	1.3973	0.8996	0.9013	0.9	6.9755	15	2	11.3031	2.2086
0.2384	34.0	1020	1.4041	0.8988	0.9001	0.899	6.9865	15	2	11.2785	2.3313
0.2369	35.0	1050	1.4044	0.8986	0.9005	0.8991	6.9595	15	2	11.2785	1.9632
0.2307	36.0	1080	1.4109	0.8985	0.9	0.8988	6.9436	15	2	11.2712	2.0859
0.2285	37.0	1110	1.4118	0.8987	0.8995	0.8987	6.9067	15	2	11.2037	1.7178
0.2294	38.0	1140	1.4171	0.8992	0.9007	0.8995	6.9436	15	2	11.2712	1.9632
0.2294	39.0	1170	1.4188	0.8988	0.9008	0.8994	6.9656	15	2	11.3006	2.2086
0.2261	40.0	1200	1.4183	0.8989	0.9008	0.8994	6.9571	15	2	11.2896	2.0859

Framework versions

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