gemma-simpo-reproduction

This model is a fine-tuned version of google/gemma-2-9b-it on the mlfoundations-dev/gemma2-ultrafeedback-armorm dataset. It achieves the following results on the evaluation set:

• Loss: 3.0558
• Rewards/chosen: -17.0597
• Rewards/rejected: -21.9498
• Rewards/accuracies: 0.7584
• Rewards/margins: 4.8901
• Logps/rejected: -2.1950
• Logps/chosen: -1.7060
• Logits/rejected: -18.1137
• Logits/chosen: -18.2041

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: 1
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 32
• total_train_batch_size: 128
• total_eval_batch_size: 4
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 1.0

Training results

Framework versions

• Transformers 4.45.2
• Pytorch 2.2.0+cu121
• Datasets 3.0.0
• Tokenizers 0.20.1