QA-Qwen3-8B-4164

This model is a fine-tuned version of Qwen/Qwen3-8B on the saiteki-kai/BeaverTails-it dataset. It achieves the following results on the evaluation set:

• Loss: 0.0780
• Accuracy: 0.6972
• Macro F1: 0.6342
• Macro Precision: 0.7694
• Macro Recall: 0.5759
• Micro F1: 0.7527
• Micro Precision: 0.8153
• Micro Recall: 0.6990
• Flagged/accuracy: 0.8542
• Flagged/precision: 0.9105
• Flagged/recall: 0.8185
• Flagged/f1: 0.8620

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: 3.37220786752157e-07
• train_batch_size: 8
• eval_batch_size: 32
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• total_train_batch_size: 32
• total_eval_batch_size: 128
• optimizer: Use 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_ratio: 0.1
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Macro F1	Macro Precision	Macro Recall	Micro F1	Micro Precision	Micro Recall	Flagged/accuracy	Flagged/precision	Flagged/recall	Flagged/f1
0.0714	1.0	16907	0.0802	0.6923	0.6343	0.7452	0.5775	0.7445	0.8127	0.6869	0.8472	0.9047	0.8109	0.8552
0.0802	2.0	33814	0.0780	0.6972	0.6341	0.7688	0.5760	0.7528	0.8153	0.6992	0.8540	0.9103	0.8183	0.8619
0.0733	3.0	50721	0.0787	0.6951	0.6443	0.7459	0.5936	0.7524	0.8061	0.7053	0.8535	0.9046	0.8236	0.8622

Framework versions

• Transformers 4.51.3
• Pytorch 2.7.0+cu118
• Datasets 3.5.1
• Tokenizers 0.21.1