bert-small-finer-longer

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

• Loss: 1.4264

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: 5e-05
• train_batch_size: 8
• eval_batch_size: 64
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.49	500	1.6683
1.5941	0.97	1000	1.6569
1.5941	1.46	1500	1.6436
1.5605	1.94	2000	1.6173
1.5605	2.43	2500	1.6073
1.5297	2.91	3000	1.6001
1.5297	3.4	3500	1.5815
1.5022	3.89	4000	1.5756
1.5022	4.37	4500	1.5568
1.4753	4.86	5000	1.5458
1.4753	5.34	5500	1.5399
1.4537	5.83	6000	1.5273
1.4537	6.32	6500	1.5192
1.433	6.8	7000	1.5099
1.433	7.29	7500	1.5083
1.4169	7.77	8000	1.4957
1.4169	8.26	8500	1.4914
1.3982	8.75	9000	1.4859
1.3982	9.23	9500	1.4697
1.3877	9.72	10000	1.4711
1.3877	10.2	10500	1.4608
1.3729	10.69	11000	1.4583
1.3729	11.18	11500	1.4513
1.3627	11.66	12000	1.4498
1.3627	12.15	12500	1.4396
1.357	12.63	13000	1.4415
1.357	13.12	13500	1.4347
1.3484	13.61	14000	1.4316
1.3484	14.09	14500	1.4319
1.3442	14.58	15000	1.4268
1.3442	15.06	15500	1.4293
1.3387	15.55	16000	1.4217
1.3387	16.03	16500	1.4241
1.3358	16.52	17000	1.4250
1.3358	17.01	17500	1.4196
1.3344	17.49	18000	1.4193
1.3344	17.98	18500	1.4200
1.3274	18.46	19000	1.4250
1.3274	18.95	19500	1.4168
1.3348	19.44	20000	1.4164
1.3348	19.92	20500	1.4264

Framework versions

• Transformers 4.18.0
• Pytorch 1.12.1+cu113
• Datasets 2.4.0
• Tokenizers 0.12.1