qwen3-8b-base-slic-hf-ultrafeedback-4xh200-batch-128-20260422-131855

• Loss: 330.7802
• Rewards/chosen: -250.5781
• Rewards/rejected: -238.5265
• Rewards/accuracies: 0.4945
• Rewards/margins: -12.0516
• Logps/chosen: -250.5781
• Logps/rejected: -238.5265
• Slic/rank Loss: 80.2021
• Slic/ce Loss: 250.5781
• Logits/chosen: 2.4579
• Logits/rejected: 2.4852

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

• learning_rate: 5e-07
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 8
• total_train_batch_size: 128
• total_eval_batch_size: 16
• 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_ratio: 0.1
• num_epochs: 1

Training results

Framework versions

• Transformers 4.51.0
• Pytorch 2.3.1+cu121
• Datasets 2.21.0
• Tokenizers 0.21.4