byt5-base-ko-1e4-jit-20epochs-a100-jamoparams-wu2000-rm1-bs2-ga64-eb128-distilled-small-7

• Loss: 1.0463

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

• learning_rate: 0.0005
• train_batch_size: 2
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 64
• total_train_batch_size: 128
• optimizer: Use adamw_torch with betas=(0.9,0.98) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: inverse_sqrt
• lr_scheduler_warmup_steps: 2000
• num_epochs: 40
• label_smoothing_factor: 0.1

Training results

Framework versions

• Transformers 4.57.3
• Pytorch 2.9.1+cu128
• Datasets 4.4.1
• Tokenizers 0.22.1