ner_based_ChineseModernBert_without_phone

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

• Loss: 0.0239
• Precision: 0.9365
• Recall: 0.9401
• F1: 0.9383
• Accuracy: 0.9960

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: 2e-05
• train_batch_size: 64
• eval_batch_size: 64
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 512
• 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: cosine
• lr_scheduler_warmup_steps: 20
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	1.0	163	0.0310	0.8354	0.8581	0.8466	0.9907
No log	2.0	326	0.0202	0.8915	0.9244	0.9077	0.9942
No log	3.0	489	0.0174	0.9211	0.9297	0.9254	0.9953
0.0784	4.0	652	0.0167	0.9226	0.9374	0.9300	0.9956
0.0784	5.0	815	0.0182	0.9305	0.9434	0.9369	0.9959
0.0784	6.0	978	0.0193	0.9329	0.9438	0.9383	0.9960
0.0046	7.0	1141	0.0220	0.9333	0.9430	0.9381	0.9960
0.0046	8.0	1304	0.0221	0.9328	0.9437	0.9382	0.9959
0.0046	9.0	1467	0.0235	0.9320	0.9452	0.9385	0.9960
0.0011	10.0	1630	0.0239	0.9365	0.9401	0.9383	0.9960

Framework versions

• Transformers 4.54.0
• Pytorch 2.7.0+cu128
• Datasets 4.0.0
• Tokenizers 0.21.4