text_shortening_model_v72

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

• Loss: 1.6295
• Bert precision: 0.9015
• Bert recall: 0.9003
• Bert f1-score: 0.9004
• Average word count: 6.4845
• Max word count: 16
• Min word count: 2
• Average token count: 10.5656
• % shortened texts with length > 12: 1.1011

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: 0.0005
• 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
1.6981	1.0	37	1.2099	0.8879	0.8868	0.8868	6.5786	15	1	10.3994	0.8008
1.1993	2.0	74	1.1320	0.8939	0.89	0.8914	6.3013	16	2	10.2663	0.9009
1.0205	3.0	111	1.1073	0.8929	0.8931	0.8925	6.6507	16	2	10.7057	1.6016
0.8912	4.0	148	1.0787	0.8967	0.8966	0.8962	6.5896	16	2	10.5926	1.6016
0.8027	5.0	185	1.1123	0.8991	0.8959	0.897	6.3994	16	2	10.4164	1.1011
0.7251	6.0	222	1.1148	0.8983	0.8941	0.8957	6.3013	16	2	10.3333	1.3013
0.6534	7.0	259	1.1348	0.8993	0.8931	0.8957	6.2332	16	2	10.2012	1.2012
0.5895	8.0	296	1.1537	0.8982	0.8959	0.8966	6.4945	16	2	10.4995	1.6016
0.5483	9.0	333	1.1656	0.901	0.8978	0.899	6.4184	16	2	10.4505	1.7017
0.5117	10.0	370	1.1919	0.8977	0.896	0.8964	6.4565	15	2	10.5696	1.1011
0.4639	11.0	407	1.2106	0.8999	0.8956	0.8973	6.2653	15	2	10.2943	1.001
0.4267	12.0	444	1.2419	0.8975	0.8958	0.8962	6.4625	17	2	10.5115	1.7017
0.4069	13.0	481	1.2583	0.9023	0.8964	0.8988	6.1812	15	2	10.1942	0.9009
0.3775	14.0	518	1.2887	0.8991	0.8982	0.8982	6.4384	15	2	10.5676	1.5015
0.3495	15.0	555	1.3282	0.9015	0.8984	0.8995	6.3604	15	2	10.4895	0.9009
0.3281	16.0	592	1.3276	0.9012	0.8973	0.8988	6.2753	15	2	10.3413	0.5005
0.3083	17.0	629	1.3539	0.9007	0.8979	0.8989	6.3504	16	2	10.3874	1.6016
0.2906	18.0	666	1.3720	0.9006	0.8986	0.8992	6.4204	14	2	10.4785	1.2012
0.2793	19.0	703	1.4130	0.8997	0.8986	0.8987	6.4374	16	2	10.5345	1.5015
0.2656	20.0	740	1.4376	0.9026	0.8986	0.9002	6.2843	16	2	10.3834	1.2012
0.2399	21.0	777	1.4429	0.901	0.8997	0.8999	6.4545	16	2	10.5516	1.5015
0.2316	22.0	814	1.4807	0.899	0.8987	0.8983	6.4975	16	2	10.6667	1.3013
0.2237	23.0	851	1.4941	0.9002	0.8974	0.8983	6.3363	15	2	10.4484	0.9009
0.2079	24.0	888	1.5101	0.9011	0.8982	0.8992	6.3443	16	2	10.4104	1.2012
0.2007	25.0	925	1.5176	0.8991	0.8983	0.8982	6.5065	16	2	10.6216	1.001
0.1952	26.0	962	1.5253	0.9005	0.8979	0.8987	6.3934	15	2	10.4835	1.1011
0.1901	27.0	999	1.5440	0.9007	0.8985	0.8991	6.3904	16	2	10.5185	0.8008
0.1838	28.0	1036	1.5540	0.9008	0.9002	0.9	6.4985	16	2	10.6176	1.3013
0.1773	29.0	1073	1.5576	0.9013	0.9001	0.9003	6.4835	16	2	10.5866	1.3013
0.1692	30.0	1110	1.5746	0.9012	0.9003	0.9003	6.4895	16	2	10.6176	1.5015
0.163	31.0	1147	1.5844	0.9014	0.9	0.9002	6.4655	16	2	10.5756	1.3013
0.1587	32.0	1184	1.6071	0.9008	0.8997	0.8998	6.4615	16	2	10.6076	0.9009
0.156	33.0	1221	1.6166	0.9006	0.8998	0.8997	6.4945	16	2	10.6166	1.2012
0.1546	34.0	1258	1.6099	0.9011	0.8987	0.8994	6.3834	13	2	10.4965	0.9009
0.1472	35.0	1295	1.6167	0.9018	0.8992	0.9001	6.3974	14	2	10.4665	1.001
0.1472	36.0	1332	1.6271	0.9006	0.9	0.8998	6.5185	16	2	10.6216	1.5015
0.1452	37.0	1369	1.6226	0.9023	0.9007	0.901	6.4595	16	2	10.5485	1.4014
0.1415	38.0	1406	1.6221	0.9015	0.9006	0.9006	6.5005	16	2	10.5846	1.4014
0.1398	39.0	1443	1.6272	0.9012	0.9002	0.9003	6.5025	16	2	10.5866	1.2012
0.14	40.0	1480	1.6295	0.9015	0.9003	0.9004	6.4845	16	2	10.5656	1.1011

Framework versions

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