vumichien/ner-jp-gliner

• Precision: 84.79%
• Recall: 75.04%
• F1 score: 79.62%

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

• num_steps: 30000
• train_batch_size: 8
• eval_every: 3000
• warmup_ratio: 0.1
• scheduler_type: "cosine"
• loss_alpha: -1
• loss_gamma: 0
• label_smoothing: 0
• loss_reduction: "sum"
• lr_encoder: 1e-5
• lr_others: 5e-5
• weight_decay_encoder: 0.01
• weight_decay_other: 0.01

Training results