bert-base-cased-cv-studio_name-medium

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

• Loss: 2.3310
• F1 Micro: 0.6388
• F1 Macro: 0.5001

Model description

Predicts a studio name based on a CV text

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 20
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1 Micro	F1 Macro	Precision Micro	Recall Micro
1.4139	0.98	1000	1.3831	0.6039	0.6039	0.4188	0.6039	0.6039
1.1561	1.96	2000	1.2386	0.6554	0.6554	0.4743	0.6554	0.6554
0.9183	2.93	3000	1.2201	0.6576	0.6576	0.5011	0.6576	0.6576
0.677	3.91	4000	1.3478	0.6442	0.6442	0.5206	0.6442	0.6442
0.4857	4.89	5000	1.4765	0.6393	0.6393	0.5215	0.6393	0.6393
0.3318	5.87	6000	1.6924	0.6442	0.6442	0.5024	0.6442	0.6442
0.2273	6.84	7000	1.8645	0.6444	0.6444	0.5060	0.6444	0.6444
0.1396	7.82	8000	2.1143	0.6381	0.6381	0.5181	0.6381	0.6381
0.0841	8.8	9000	2.2699	0.6359	0.6359	0.5065	0.6359	0.6359
0.0598	9.78	10000	2.3310	0.6388	0.6388	0.5001	0.6388	0.6388

Framework versions

• Transformers 4.19.0
• Pytorch 1.8.2+cu111
• Datasets 1.6.2
• Tokenizers 0.12.1