modernbert-base-multi-head

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

• Loss: 0.6827
• F1 Macro: 0.4188

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-06
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 16
• 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_ratio: 0.05
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	F1 Macro
0.5788	1.0	1534	0.5783	0.0
0.5675	2.0	3068	0.5714	0.1304
0.5424	3.0	4602	0.5386	0.1952
0.516	4.0	6136	0.5493	0.2382
0.4922	5.0	7670	0.6015	0.2394
0.4744	6.0	9204	0.5404	0.3192
0.4494	7.0	10738	0.5746	0.3733
0.4353	8.0	12272	0.5689	0.3396
0.4147	9.0	13806	0.5695	0.3608
0.4006	10.0	15340	0.6441	0.3871
0.3843	11.0	16874	0.5938	0.3888
0.3763	12.0	18408	0.6159	0.4054
0.3606	13.0	19942	0.5831	0.4042
0.3478	14.0	21476	0.6116	0.4056
0.3447	15.0	23010	0.6169	0.3999
0.3215	16.0	24544	0.6441	0.4150
0.3196	17.0	26078	0.6530	0.4129
0.3184	18.0	27612	0.6634	0.4192
0.3065	19.0	29146	0.6791	0.4192
0.298	20.0	30680	0.6827	0.4188

Framework versions

• Transformers 4.53.1
• Pytorch 2.6.0+cu124
• Datasets 2.14.4
• Tokenizers 0.21.2