simpo-oh-dcft-v3.1-llama-3.3-70b

This model is a fine-tuned version of mlfoundations-dev/oh-dcft-v3.1-llama-3.3-70b on the mlfoundations-dev/gemma2-ultrafeedback-armorm dataset. It achieves the following results on the evaluation set:

• Loss: 2.2941
• Rewards/chosen: -35.1648
• Rewards/rejected: -43.3712
• Rewards/accuracies: 0.8115
• Rewards/margins: 8.2064
• Logps/chosen: -3.5165
• Logps/rejected: -4.3371
• Logits/chosen: -1.2001
• Logits/rejected: -1.1966

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: 8e-07
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• 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: 1.0

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/chosen	Logps/rejected	Logits/chosen	Logits/rejected
2.4511	0.9997	442	2.2941	-35.1648	-43.3712	0.8115	8.2064	-3.5165	-4.3371	-1.2001	-1.1966

Framework versions

• Transformers 4.46.1
• Pytorch 2.3.0
• Datasets 3.1.0
• Tokenizers 0.20.3