tibetan-CS-detector_mbert-tibetan-continual-wylie_all_data_no_labels_no_partial

This model is a fine-tuned version of OMRIDRORI/mbert-tibetan-continual-wylie-final on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 17.6968
• Accuracy: 0.9266
• Switch Precision: 0.5017
• Switch Recall: 0.9091
• Switch F1: 0.6466
• True Switches: 165
• Pred Switches: 299
• Exact Matches: 144
• Proximity Matches: 6
• To Auto Precision: 0.6372
• To Auto Recall: 0.9
• To Allo Precision: 0.4194
• To Allo Recall: 0.9176
• True To Auto: 80
• True To Allo: 85
• Matched To Auto: 72
• Matched To Allo: 78

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: 1e-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 200
• num_epochs: 35
• mixed_precision_training: Native AMP
• label_smoothing_factor: 0.05

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Switch Precision	Switch Recall	Switch F1	True Switches	Pred Switches	Exact Matches	Proximity Matches	To Auto Precision	To Auto Recall	To Allo Precision	To Allo Recall	True To Auto	True To Allo	Matched To Auto	Matched To Allo
9.3149	1.5789	30	3.6306	0.5806	0.0	0.0	0.0	165	2	0	0	0.0	0.0	0.0	0.0	80	85	0	0
4.2352	3.1579	60	3.0508	0.7829	0.0	0.0	0.0	165	1	0	0	0.0	0.0	0.0	0.0	80	85	0	0
3.8336	4.7368	90	2.6432	0.7946	0.7963	0.2606	0.3927	165	54	42	1	0.82	0.5125	0.5	0.0235	80	85	41	2
7.8192	6.3158	120	3.5075	0.7951	0.4525	0.4909	0.4709	165	179	79	2	0.67	0.8375	0.1772	0.1647	80	85	67	14
13.6795	7.8947	150	3.5598	0.8145	0.5064	0.4788	0.4922	165	156	77	2	0.75	0.7875	0.2222	0.1882	80	85	63	16
16.8577	9.4737	180	2.6579	0.8798	0.5753	0.5091	0.5402	165	146	83	1	0.6731	0.875	0.3333	0.1647	80	85	70	14
6.7643	11.0526	210	3.0767	0.8965	0.5067	0.6848	0.5825	165	223	111	2	0.6698	0.8875	0.3590	0.4941	80	85	71	42
2.2893	12.6316	240	2.4180	0.9070	0.5642	0.7455	0.6423	165	218	120	3	0.6372	0.9	0.4857	0.6	80	85	72	51
5.0292	14.2105	270	17.3441	0.9111	0.5292	0.8242	0.6445	165	257	131	5	0.6261	0.9	0.4507	0.7529	80	85	72	64
5.1075	15.7895	300	17.8003	0.9127	0.4768	0.8727	0.6167	165	302	140	4	0.6050	0.9	0.3934	0.8471	80	85	72	72
1.7376	17.3684	330	18.0361	0.9128	0.4719	0.8667	0.6111	165	303	137	6	0.6207	0.9	0.3797	0.8353	80	85	72	71
1.7352	18.9474	360	32.0882	0.9217	0.4727	0.8909	0.6176	165	311	141	6	0.5581	0.9	0.4121	0.8824	80	85	72	75
3.8194	20.5263	390	30.3409	0.9216	0.5620	0.8788	0.6856	165	258	142	3	0.6827	0.8875	0.4805	0.8706	80	85	71	74
4.0106	22.1053	420	48.1283	0.9250	0.4654	0.8970	0.6128	165	318	140	8	0.6154	0.9	0.3781	0.8941	80	85	72	76
3.2054	23.6842	450	17.7213	0.9222	0.5121	0.8970	0.6520	165	289	143	5	0.6486	0.9	0.4270	0.8941	80	85	72	76
6.8048	25.2632	480	16.7449	0.9240	0.5068	0.9030	0.6492	165	294	142	7	0.6261	0.9	0.4302	0.9059	80	85	72	77
1.2021	26.8421	510	43.0137	0.9240	0.4760	0.9030	0.6234	165	313	142	7	0.6102	0.9	0.3949	0.9059	80	85	72	77
1.3674	28.4211	540	41.8534	0.9249	0.4715	0.9030	0.6195	165	316	142	7	0.6261	0.9	0.3831	0.9059	80	85	72	77
1.2001	30.0	570	42.0665	0.9260	0.4717	0.9091	0.6211	165	318	143	7	0.6372	0.9	0.3805	0.9176	80	85	72	78
1.1834	31.5789	600	32.0717	0.9214	0.4792	0.9091	0.6276	165	313	143	7	0.6102	0.9	0.4	0.9176	80	85	72	78
1.1969	33.1579	630	17.6968	0.9266	0.5017	0.9091	0.6466	165	299	144	6	0.6372	0.9	0.4194	0.9176	80	85	72	78

Framework versions

• Transformers 4.46.3
• Pytorch 2.4.1+cu121
• Datasets 2.0.0
• Tokenizers 0.20.3