tinybert_train_book_v2

This model is a fine-tuned version of distilbert-base-uncased on the gokulsrinivasagan/processed_book_corpus-ld dataset. It achieves the following results on the evaluation set:

• Loss: 1.4246
• Accuracy: 0.6906

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: 0.0001
• train_batch_size: 160
• eval_batch_size: 160
• seed: 10
• optimizer: Use OptimizerNames.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: 10000
• num_epochs: 25

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
5.6396	0.7025	10000	5.4784	0.1645
5.5513	1.4051	20000	5.4083	0.1660
2.9289	2.1076	30000	2.5488	0.5214
2.5686	2.8102	40000	2.2269	0.5650
2.3542	3.5127	50000	2.0268	0.5945
2.2044	4.2153	60000	1.8924	0.6152
2.1102	4.9178	70000	1.8025	0.6296
2.0378	5.6203	80000	1.7395	0.6393
1.9881	6.3229	90000	1.6945	0.6467
1.9469	7.0254	100000	1.6552	0.6526
1.919	7.7280	110000	1.6279	0.6572
1.8899	8.4305	120000	1.6012	0.6613
1.8634	9.1331	130000	1.5849	0.6641
1.8481	9.8356	140000	1.5630	0.6679
1.8315	10.5381	150000	1.5476	0.6703
1.8209	11.2407	160000	1.5394	0.6716
1.8038	11.9432	170000	1.5250	0.6740
1.7932	12.6458	180000	1.5107	0.6761
1.7826	13.3483	190000	1.5058	0.6770
1.7736	14.0509	200000	1.4952	0.6785
1.7635	14.7534	210000	1.4862	0.6799
1.7549	15.4560	220000	1.4782	0.6815
1.7497	16.1585	230000	1.4737	0.6821
1.7407	16.8610	240000	1.4675	0.6832
1.7334	17.5636	250000	1.4612	0.6843
1.7288	18.2661	260000	1.4568	0.6849
1.7265	18.9687	270000	1.4519	0.6858
1.7179	19.6712	280000	1.4469	0.6868
1.7156	20.3738	290000	1.4428	0.6873
1.7086	21.0763	300000	1.4389	0.6882
1.7064	21.7788	310000	1.4334	0.6892
1.7009	22.4814	320000	1.4309	0.6895
1.6989	23.1839	330000	1.4292	0.6896
1.6971	23.8865	340000	1.4262	0.6902
1.6958	24.5890	350000	1.4239	0.6906

Framework versions

• Transformers 4.46.1
• Pytorch 2.2.0+cu121
• Datasets 3.1.0
• Tokenizers 0.20.1