finetuned-BART-UK-financial-summarization-multiple-references

This model is a fine-tuned version of facebook/bart-base on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.8651
• Rouge1: 30.2768
• Rouge2: 16.5597
• Rougel: 23.9264
• Rougelsum: 29.9211
• Gen Len: 114.7128

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: 1
• eval_batch_size: 4
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 4.0

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len
3.0082	0.1	1000	2.4917	13.4643	8.5691	12.4617	13.3171	20.0
2.4695	0.2	2000	2.3554	12.3599	8.0996	11.4208	12.2321	20.0
2.2945	0.3	3000	2.2914	12.6496	8.3592	11.7265	12.5178	19.8696
2.4193	0.41	4000	2.2373	13.1184	8.7248	12.2447	13.019	20.0
2.2386	0.51	5000	2.2263	11.7267	7.5929	10.8553	11.5917	18.264
2.2384	0.61	6000	2.2009	13.6181	9.1253	12.6438	13.4864	20.0
2.1491	0.71	7000	2.1802	12.7595	8.2607	11.8995	12.6552	20.0
2.1001	0.81	8000	2.1458	12.6461	8.2799	11.7226	12.5094	19.9408
2.0731	0.91	9000	2.0940	13.098	8.5378	12.0868	13.0062	19.8944
2.0307	1.01	10000	2.0865	12.9732	8.5554	12.0287	12.8468	19.9712
1.8416	1.12	11000	2.0670	13.1366	8.6049	12.2127	13.0122	19.9184
1.9564	1.22	12000	2.0494	12.664	8.4708	11.8496	12.5582	19.9856
1.8846	1.32	13000	2.0231	13.3412	8.9207	12.4697	13.2275	19.9904
1.8347	1.42	14000	2.0383	12.5847	8.3611	11.6671	12.4601	20.0
1.8773	1.52	15000	1.9892	12.6655	8.2128	11.7945	12.5593	19.9592
1.8015	1.62	16000	1.9959	13.0598	8.7329	12.1663	12.9552	19.9736
1.8119	1.72	17000	1.9758	12.7897	8.4912	11.8756	12.6557	20.0
1.8025	1.83	18000	1.9764	13.2535	8.8228	12.3222	13.1407	19.948
1.8417	1.93	19000	1.9613	12.5747	8.3244	11.6604	12.4882	19.9712
1.6609	2.03	20000	1.9648	12.7294	8.4078	11.8369	12.6136	20.0
1.5885	2.13	21000	1.9464	13.2406	8.9054	12.4006	13.1491	19.9688
1.6082	2.23	22000	1.9442	13.0069	8.6038	12.1138	12.9006	20.0
1.5773	2.33	23000	1.9340	12.6373	8.3122	11.7935	12.5309	20.0
1.6535	2.43	24000	1.9432	13.1702	8.8258	12.2924	13.0442	19.9712
1.5923	2.53	25000	1.9226	13.2556	8.9101	12.3433	13.136	19.9648
1.4997	2.64	26000	1.9128	13.0711	8.6794	12.1325	12.9455	19.9904
1.574	2.74	27000	1.9093	13.1309	8.7958	12.2906	13.0341	19.9904
1.6254	2.84	28000	1.8991	12.8586	8.5662	12.0449	12.7786	19.9832
1.6427	2.94	29000	1.8905	12.7339	8.5642	11.9439	12.6131	19.996
1.3442	3.04	30000	1.9000	12.818	8.5827	11.9443	12.6917	20.0
1.466	3.14	31000	1.8990	12.9079	8.6233	12.0834	12.7858	19.9864
1.4822	3.24	32000	1.8937	13.2885	9.061	12.4811	13.1978	19.952
1.4399	3.35	33000	1.8927	13.0858	8.8237	12.2749	12.974	19.9624
1.3888	3.45	34000	1.8776	12.955	8.7827	12.1243	12.8524	19.9864
1.4031	3.55	35000	1.8816	12.8846	8.6634	11.9851	12.775	19.9904
1.4458	3.65	36000	1.8732	12.8441	8.5444	11.9473	12.728	19.9784
1.401	3.75	37000	1.8734	12.8671	8.616	12.0224	12.7432	19.9568
1.3798	3.85	38000	1.8664	12.7505	8.5224	11.8733	12.6269	19.9904
1.4413	3.95	39000	1.8656	12.8364	8.609	11.9612	12.7237	19.9904

Framework versions

• Transformers 4.16.0.dev0
• Pytorch 1.7.0
• Datasets 1.17.0
• Tokenizers 0.11.0