bert_12_layer_model_v3_complete_training_48

This model is a fine-tuned version of on the gokuls/wiki_book_corpus_complete_processed_bert_dataset dataset. It achieves the following results on the evaluation set:

• Loss: 3.2767
• Accuracy: 0.4430

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: 48
• eval_batch_size: 48
• seed: 10
• distributed_type: multi-GPU
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 10000
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
6.5785	0.08	10000	6.5368	0.1282
6.3475	0.16	20000	6.3341	0.1409
6.2664	0.25	30000	6.2525	0.1446
6.2198	0.33	40000	6.2002	0.1474
6.1403	0.41	50000	6.1243	0.1497
6.0674	0.49	60000	6.0389	0.1520
5.9921	0.57	70000	5.9584	0.1548
5.9337	0.66	80000	5.8929	0.1577
5.8269	0.74	90000	5.7654	0.1685
5.5053	0.82	100000	5.3216	0.2274
4.8237	0.9	110000	4.6350	0.3054
4.3302	0.98	120000	4.1723	0.3496
4.0721	1.07	130000	3.9373	0.3741
3.8896	1.15	140000	3.7763	0.3903
3.7842	1.23	150000	3.6628	0.4018
3.6871	1.31	160000	3.5734	0.4111
3.6099	1.39	170000	3.4986	0.4188
3.5585	1.47	180000	3.4324	0.4259
3.4922	1.56	190000	3.3767	0.4316
3.4342	1.64	200000	3.3234	0.4374
3.3799	1.72	210000	3.2770	0.4429

Framework versions

• Transformers 4.33.3
• Pytorch 1.14.0a0+410ce96
• Datasets 2.14.5
• Tokenizers 0.13.3