library_name, base_model, tags, datasets, model-index

library_name	base_model	tags	datasets	model-index
transformers	W-61/qwen3-8b-base-sft-ultrachat-4xh200-batch-128	alignment-handbook beta-dpo generated_from_trainer	HuggingFaceH4/ultrafeedback_binarized	name results qwen3-8b-base-beta-dpo-ultrafeedback-4xh200-batch-128-20260423-040315

qwen3-8b-base-beta-dpo-ultrafeedback-4xh200-batch-128-20260423-040315

This model is a fine-tuned version of W-61/qwen3-8b-base-sft-ultrachat-4xh200-batch-128 on the HuggingFaceH4/ultrafeedback_binarized dataset. It achieves the following results on the evaluation set:

• Loss: 0.5897
• Beta Dpo/gap Mean: 23.0136
• Beta Dpo/gap Std: 39.9127
• Beta Dpo/beta Used Raw: 0.0146
• Beta Dpo/beta Used: 0.0335
• Beta Dpo/mask Keep Frac: 1.0
• Logits/chosen: 1.5098
• Logits/rejected: 1.5463

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: 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

Training Loss	Epoch	Step	Validation Loss	Beta Dpo/gap Mean	Beta Dpo/gap Std	Beta Dpo/beta Used Raw	Beta Dpo/beta Used	Beta Dpo/mask Keep Frac	Logits/chosen	Logits/rejected
4.6811	0.4188	200	0.5822	17.3495	36.2958	0.0115	0.0273	1.0	1.4601	1.4735
4.2926	0.8377	400	0.5897	23.0136	39.9127	0.0146	0.0335	1.0	1.5098	1.5463

Framework versions

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