bert-base-uncased-bert-base-uncased-mc-weight0-epoch15

This model is a fine-tuned version of on the None dataset. It achieves the following results on the evaluation set:

• Loss: 4.3651
• Cls loss: 2.9223
• Lm loss: 4.3649
• Cls Accuracy: 0.0248
• Cls F1: 0.0057
• Cls Precision: 0.0061
• Cls Recall: 0.0248
• Perplexity: 78.64

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: 2
• eval_batch_size: 2
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 15

Training results

Training Loss	Epoch	Step	Validation Loss	Cls loss	Lm loss	Cls Accuracy	Cls F1	Cls Precision	Cls Recall	Perplexity
4.8711	1.0	3470	4.5156	2.9252	4.5155	0.0213	0.0047	0.0042	0.0213	91.42
4.483	2.0	6940	4.4193	2.9248	4.4191	0.0219	0.0048	0.0042	0.0219	83.02
4.3345	3.0	10410	4.3684	2.9244	4.3682	0.0219	0.0048	0.0042	0.0219	78.91
4.2266	4.0	13880	4.3445	2.9241	4.3443	0.0225	0.0049	0.0043	0.0225	77.04
4.1388	5.0	17350	4.3260	2.9237	4.3258	0.0231	0.0050	0.0044	0.0231	75.63
4.0644	6.0	20820	4.3299	2.9234	4.3297	0.0231	0.0050	0.0044	0.0231	75.92
3.999	7.0	24290	4.3278	2.9232	4.3276	0.0231	0.0059	0.0061	0.0231	75.76
3.9426	8.0	27760	4.3269	2.9230	4.3267	0.0231	0.0059	0.0061	0.0231	75.70
3.8929	9.0	31230	4.3324	2.9228	4.3322	0.0248	0.0061	0.0062	0.0248	76.11
3.8488	10.0	34700	4.3382	2.9227	4.3380	0.0248	0.0061	0.0064	0.0248	76.55
3.8116	11.0	38170	4.3461	2.9225	4.3459	0.0242	0.0057	0.0061	0.0242	77.16
3.7791	12.0	41640	4.3537	2.9224	4.3535	0.0248	0.0057	0.0061	0.0248	77.75
3.7532	13.0	45110	4.3593	2.9223	4.3591	0.0248	0.0057	0.0061	0.0248	78.19
3.7321	14.0	48580	4.3588	2.9223	4.3586	0.0248	0.0057	0.0061	0.0248	78.15
3.7182	15.0	52050	4.3651	2.9223	4.3649	0.0248	0.0057	0.0061	0.0248	78.64

Framework versions

• Transformers 4.21.2
• Pytorch 1.12.1
• Datasets 2.4.0
• Tokenizers 0.12.1