retriva-bert-preference-classifier

This model is a fine-tuned version of retrieva-jp/bert-1.3b on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.4714
• Accuracy: 0.737
• Precision: 0.7423
• Recall: 0.726
• F1: 0.7341

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: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• training_steps: 1000

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
0.6438	0.0080	100	0.6116	0.663	0.8721	0.382	0.5313
0.5113	0.0160	200	0.5442	0.699	0.6736	0.772	0.7195
0.4512	0.0240	300	0.5119	0.717	0.8359	0.54	0.6561
0.3916	0.0321	400	0.4936	0.702	0.7295	0.642	0.6830
0.3806	0.0401	500	0.4763	0.715	0.7708	0.612	0.6823
0.3581	0.0481	600	0.4597	0.754	0.75	0.762	0.7560
0.3308	0.0561	700	0.4690	0.742	0.7738	0.684	0.7261
0.3458	0.0641	800	0.4703	0.737	0.7423	0.726	0.7341
0.3475	0.0721	900	0.4728	0.737	0.7495	0.712	0.7303
0.3435	0.0801	1000	0.4714	0.737	0.7423	0.726	0.7341

Evaluation on test split

Framework versions

• Transformers 4.43.1
• Pytorch 2.3.1+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1