fine_tuned_xlm-roberta_3April

This model is a fine-tuned version of ancs21/xlm-roberta-large-vi-qa on the None dataset. It achieves the following results on the evaluation set:

• Best F1: 76.1698
• Loss: 3.3902
• Exact: 39.0997
• F1: 56.5332
• Total: 3821
• Hasans Exact: 56.1628
• Hasans F1: 81.2715
• Hasans Total: 2653
• Noans Exact: 0.3425
• Noans F1: 0.3425
• Noans Total: 1168
• Best Exact: 60.8479
• Best Exact Thresh: 0.5169
• Best F1 Thresh: 0.6480

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: 2e-05
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Best F1	Validation Loss	Exact	F1	Total	Hasans Exact	Hasans F1	Hasans Total	Noans Exact	Noans F1	Noans Total	Best Exact	Best Exact Thresh	Best F1 Thresh
0.6591	0.24	1000	69.8185	1.2645	36.5611	54.8584	3821	52.6574	79.0101	2653	0.0	0.0	1168	55.2473	0.8809	0.9575
0.576	0.47	2000	72.9418	1.2144	37.6865	55.8902	3821	54.2405	80.4585	2653	0.0856	0.0856	1168	57.2363	0.9300	0.9414
0.5355	0.71	3000	72.9946	1.1680	39.0474	55.7491	3821	56.2382	80.2930	2653	0.0	0.0	1168	58.9375	0.9023	0.9418
0.558	0.95	4000	71.7133	1.2049	38.2361	55.4918	3821	54.7305	79.5832	2653	0.7705	0.7705	1168	57.2887	0.8935	0.9786
0.4136	1.18	5000	71.3382	1.2880	38.7071	55.5727	3821	55.6351	79.9258	2653	0.2568	0.2568	1168	57.3934	0.6526	0.8172
0.4046	1.42	6000	73.7959	1.1542	38.4978	55.9133	3821	55.1451	80.2280	2653	0.6849	0.6849	1168	58.8851	0.8676	0.9130
0.3991	1.66	7000	73.8805	1.1187	38.9950	56.2014	3821	56.1628	80.9445	2653	0.0	0.0	1168	59.5394	0.8720	0.9676
0.4062	1.9	8000	74.6655	1.0558	38.3146	55.9064	3821	55.1451	80.4818	2653	0.0856	0.0856	1168	59.4609	0.7752	0.8960
0.3246	2.13	9000	74.3864	1.3330	39.5708	56.4795	3821	56.6528	81.0058	2653	0.7705	0.7705	1168	59.9058	0.7984	0.9832
0.3016	2.37	10000	73.8462	1.3389	39.0212	56.2669	3821	56.2005	81.0388	2653	0.0	0.0	1168	59.1468	0.7780	0.8369
0.297	2.61	11000	75.1653	1.3418	39.6493	56.8063	3821	57.1052	81.8156	2653	0.0	0.0	1168	59.9320	0.8351	0.9593
0.291	2.84	12000	74.9741	1.3439	39.0997	56.5274	3821	56.3136	81.4139	2653	0.0	0.0	1168	59.3562	0.9742	0.9970
0.251	3.08	13000	75.3242	1.7083	39.5446	56.7699	3821	56.3890	81.1978	2653	1.2842	1.2842	1168	60.5339	0.9559	0.9798
0.2022	3.32	14000	75.0498	1.5213	38.7857	56.1680	3821	55.8613	80.8963	2653	0.0	0.0	1168	59.8273	0.9108	0.9803
0.2129	3.55	15000	75.1471	1.5169	38.7857	56.1669	3821	55.8236	80.8570	2653	0.0856	0.0856	1168	59.3300	0.9468	0.9995
0.2071	3.79	16000	74.6861	1.4170	39.5185	56.7831	3821	56.8790	81.7445	2653	0.0856	0.0856	1168	60.0105	0.7001	0.8805
0.2052	4.03	17000	75.7601	1.9237	39.4661	56.6037	3821	56.8036	81.4862	2653	0.0856	0.0856	1168	60.5339	0.5681	0.9653
0.1407	4.26	18000	75.5813	1.8430	38.3407	56.0982	3821	55.2205	80.7957	2653	0.0	0.0	1168	60.0105	0.5421	0.9222
0.153	4.5	19000	75.1837	1.7648	38.4716	56.0839	3821	55.3713	80.7376	2653	0.0856	0.0856	1168	60.0628	0.6387	0.9745
0.1538	4.74	20000	75.0762	1.5864	37.8697	55.5910	3821	54.5043	80.0276	2653	0.0856	0.0856	1168	59.4609	0.6394	0.9597
0.1569	4.97	21000	74.7077	1.6717	38.1837	55.8677	3821	54.9943	80.4638	2653	0.0	0.0	1168	58.8589	0.5886	0.8916
0.1091	5.21	22000	75.6475	2.0903	38.6548	56.2204	3821	55.6351	80.9341	2653	0.0856	0.0856	1168	60.1152	0.6634	0.8348
0.1107	5.45	23000	75.3914	2.2899	39.2829	56.6296	3821	56.2382	81.2220	2653	0.7705	0.7705	1168	59.6964	0.5812	0.9192
0.1147	5.69	24000	75.9043	1.9436	38.9950	56.4320	3821	56.1251	81.2388	2653	0.0856	0.0856	1168	60.5077	0.7035	0.9620
0.1097	5.92	25000	76.0467	1.8455	38.3669	55.8887	3821	55.2582	80.4940	2653	0.0	0.0	1168	60.3769	0.7807	0.9827
0.0844	6.16	26000	75.6935	2.5688	38.6810	56.3070	3821	55.6728	81.0588	2653	0.0856	0.0856	1168	59.8273	0.6209	0.9985
0.0774	6.4	27000	75.6526	2.5920	38.5763	56.1577	3821	55.5597	80.8815	2653	0.0	0.0	1168	60.0628	0.5364	0.9746
0.08	6.63	28000	75.5989	2.6070	38.9427	56.6394	3821	56.0498	81.5375	2653	0.0856	0.0856	1168	59.9581	0.5102	0.8782
0.0791	6.87	29000	75.9065	2.4939	38.6810	56.7529	3821	55.5974	81.6256	2653	0.2568	0.2568	1168	60.0890	0.7359	0.9913
0.0667	7.11	30000	75.9315	2.8027	38.6810	56.6293	3821	55.5974	81.4476	2653	0.2568	0.2568	1168	59.8011	0.6325	0.9909
0.0504	7.34	31000	75.4506	2.9797	39.0735	56.6964	3821	56.2382	81.6196	2653	0.0856	0.0856	1168	60.1413	0.5311	0.9996
0.0485	7.58	32000	75.7607	2.8116	38.7333	56.4275	3821	55.7105	81.1947	2653	0.1712	0.1712	1168	60.1413	0.5366	0.9183
0.0487	7.82	33000	75.9794	2.9559	39.1259	56.5935	3821	56.2005	81.3584	2653	0.3425	0.3425	1168	60.6386	0.4966	0.9163
0.0496	8.05	34000	75.8458	2.9892	39.2044	56.5961	3821	56.2382	81.2868	2653	0.5137	0.5137	1168	60.4292	0.4952	0.9726
0.0378	8.29	35000	75.5534	3.0456	39.0212	56.4687	3821	55.9744	81.1033	2653	0.5137	0.5137	1168	60.1152	0.4984	0.9850
0.0322	8.53	36000	75.5716	3.2717	38.9689	56.4812	3821	55.9367	81.1589	2653	0.4281	0.4281	1168	60.2460	0.5405	0.5620
0.0333	8.77	37000	76.1348	3.2260	39.0997	56.6567	3821	56.0121	81.2986	2653	0.6849	0.6849	1168	60.5601	0.6449	0.9261
0.0298	9.0	38000	76.4711	3.1900	39.3353	56.7593	3821	56.5021	81.5972	2653	0.3425	0.3425	1168	60.9526	0.5191	0.9665
0.0181	9.24	39000	76.4433	3.4204	39.3353	56.7658	3821	56.4644	81.5688	2653	0.4281	0.4281	1168	61.1882	0.5781	0.9364
0.0188	9.48	40000	75.8578	3.4566	39.2567	56.5759	3821	56.3890	81.3330	2653	0.3425	0.3425	1168	60.6909	0.4253	0.7260
0.0226	9.71	41000	76.2686	3.3418	39.4138	56.6183	3821	56.6528	81.4317	2653	0.2568	0.2568	1168	61.1882	0.5242	0.9950
0.019	9.95	42000	76.1698	3.3902	39.0997	56.5332	3821	56.1628	81.2715	2653	0.3425	0.3425	1168	60.8479	0.5169	0.6480

Framework versions

• Transformers 4.37.2
• Pytorch 1.13.1+cu117
• Datasets 2.16.1
• Tokenizers 0.15.2