albert-base-ours-run-3

This model is a fine-tuned version of albert-base-v2 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 2.4381
• Accuracy: 0.7
• Precision: 0.6579
• Recall: 0.6558
• F1: 0.6568

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: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
0.9876	1.0	200	0.9367	0.64	0.6707	0.5623	0.5425
0.7553	2.0	400	0.7936	0.66	0.6269	0.6298	0.6105
0.556	3.0	600	0.9257	0.71	0.6759	0.6504	0.6563
0.3871	4.0	800	0.9893	0.63	0.5882	0.5985	0.5876
0.2446	5.0	1000	1.1867	0.695	0.6582	0.6563	0.6566
0.1502	6.0	1200	1.6108	0.71	0.6708	0.6523	0.6585
0.1049	7.0	1400	2.4882	0.645	0.6030	0.5597	0.5649
0.0764	8.0	1600	2.0064	0.715	0.6798	0.6602	0.6651
0.032	9.0	1800	2.6447	0.655	0.5913	0.5774	0.5727
0.0177	10.0	2000	2.2460	0.675	0.6290	0.6287	0.6287
0.0153	11.0	2200	2.3537	0.69	0.6524	0.6407	0.6408
0.006	12.0	2400	2.4205	0.695	0.6582	0.6448	0.6486
0.0045	13.0	2600	2.3032	0.68	0.6394	0.6314	0.6287
0.0038	14.0	2800	2.3506	0.685	0.6388	0.6370	0.6367
0.0034	15.0	3000	2.3750	0.7	0.6590	0.6558	0.6573
0.0019	16.0	3200	2.4289	0.72	0.6819	0.6723	0.6763
0.0016	17.0	3400	2.4470	0.725	0.6892	0.6788	0.6830
0.0002	18.0	3600	2.4374	0.71	0.6700	0.6626	0.6657
0.0002	19.0	3800	2.4353	0.7	0.6579	0.6558	0.6568
0.0002	20.0	4000	2.4381	0.7	0.6579	0.6558	0.6568

Framework versions

• Transformers 4.25.1
• Pytorch 1.13.0+cu116
• Tokenizers 0.13.2