tibetan-CS-detector_mbert-tibetan-continual-wylie_MUL_loss_SAH

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: 0.0636
• Accuracy: 0.9294
• Switch Precision: 0.5571
• Switch Recall: 0.8715
• Switch F1: 0.6797
• True Switches: 179
• Pred Switches: 280
• Exact Matches: 153
• Proximity Matches: 3
• To Auto Precision: 0.5703
• To Auto Recall: 0.9012
• To Allo Precision: 0.5461
• To Allo Recall: 0.8469
• Matched To Auto: 73
• Matched To Allo: 83

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: 8
• eval_batch_size: 8
• seed: 42
• 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: 100
• num_epochs: 10
• mixed_precision_training: Native AMP

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	Matched To Auto	Matched To Allo
0.1439	0.5357	30	0.1175	0.8402	0.875	0.1955	0.3196	179	40	34	1	0.875	0.4321	0.0	0.0	35	0
0.1014	1.0714	60	0.1120	0.8391	0.2430	0.8268	0.3756	179	609	144	4	0.3854	0.9136	0.1775	0.7551	74	74
0.1042	1.6071	90	0.1031	0.9020	0.2117	0.9106	0.3435	179	770	145	18	0.4379	0.9136	0.1481	0.9082	74	89
0.0704	2.1429	120	0.0683	0.9098	0.3811	0.8771	0.5313	179	412	150	7	0.5034	0.9012	0.3146	0.8571	73	84
0.0666	2.6786	150	0.0696	0.9248	0.3594	0.8994	0.5136	179	448	153	8	0.4805	0.9136	0.2959	0.8878	74	87
0.047	3.2143	180	0.0552	0.9302	0.4561	0.8994	0.6053	179	353	153	8	0.4932	0.9012	0.4293	0.8980	73	88
0.0291	3.75	210	0.0699	0.9243	0.3327	0.9162	0.4881	179	493	149	15	0.4596	0.9136	0.2711	0.9184	74	90
0.0385	4.2857	240	0.0565	0.9288	0.4633	0.9162	0.6154	179	354	157	7	0.5102	0.9259	0.4300	0.9082	75	89
0.021	4.8214	270	0.0603	0.9191	0.4420	0.9162	0.5964	179	371	156	8	0.5172	0.9259	0.3938	0.9082	75	89
0.0198	5.3571	300	0.0588	0.9235	0.5249	0.8827	0.6583	179	301	155	3	0.5357	0.9259	0.5155	0.8469	75	83
0.0287	5.8929	330	0.0628	0.9125	0.4635	0.9218	0.6168	179	356	157	8	0.5461	0.9506	0.4093	0.8980	77	88
0.0127	6.4286	360	0.0604	0.9260	0.5267	0.8827	0.6597	179	300	155	3	0.5319	0.9259	0.5220	0.8469	75	83
0.0223	6.9643	390	0.0600	0.9315	0.5467	0.8827	0.6752	179	289	153	5	0.5725	0.9259	0.5253	0.8469	75	83
0.0154	7.5	420	0.0604	0.9220	0.5495	0.8994	0.6822	179	293	155	6	0.5952	0.9259	0.5150	0.8776	75	86
0.0135	8.0357	450	0.0614	0.9248	0.5279	0.8994	0.6653	179	305	155	6	0.5429	0.9383	0.5152	0.8673	76	85
0.0114	8.5714	480	0.0632	0.9276	0.5540	0.8883	0.6824	179	287	154	5	0.5906	0.9259	0.525	0.8571	75	84
0.0087	9.1071	510	0.0628	0.9243	0.5638	0.8883	0.6898	179	282	154	5	0.5906	0.9259	0.5419	0.8571	75	84
0.0075	9.6429	540	0.0636	0.9294	0.5571	0.8715	0.6797	179	280	153	3	0.5703	0.9012	0.5461	0.8469	73	83

Framework versions

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