emotions-entailment/distilbert-raw

distilbert-emotions-raw

This model is a fine-tuned version of distilbert/distilbert-base-uncased. It achieves the following results on the evaluation set:

• Loss: 0.3948
• F1 Micro: 0.8327
• F1 Macro: 0.7806
• Precision Micro: 0.8617
• Recall Micro: 0.8055

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 8
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 3
• total_train_batch_size: 24
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 3143
• num_epochs: 3
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	F1 Micro	F1 Macro	Precision Micro	Recall Micro
1.1966	1.0	10479	0.3998	0.8137	0.7466	0.8524	0.7784
1.1978	2.0	20958	0.3976	0.8309	0.7848	0.8462	0.8161
1.1855	3.0	31437	0.3970	0.8361	0.7916	0.8599	0.8135

Framework versions

• Transformers 5.10.2
• Pytorch 2.11.0+cu128
• Datasets 4.8.5
• Tokenizers 0.22.2