zephyr-7b-mypo3_sim-full-beta10.0-lr3e-6

This model is a fine-tuned version of alignment-handbook/zephyr-7b-sft-full on the HuggingFaceH4/ultrafeedback_binarized dataset. It achieves the following results on the evaluation set:

• Loss: 1.4142
• Rewards/chosen: 0.0604
• Rewards/rejected: -0.4122
• Rewards/accuracies: 0.6964
• Rewards/margins: 0.4726
• Logps/rejected: -1.1689
• Logps/chosen: -0.9729
• Logits/rejected: -2.6862
• Logits/chosen: -2.7202

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: 3e-06
• train_batch_size: 4
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 16
• total_train_batch_size: 256
• total_eval_batch_size: 32
• 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

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
1.4386	0.4187	100	1.4398	0.1258	-0.3325	0.6667	0.4584	-1.1609	-0.9664	-2.8124	-2.8394
1.3977	0.8375	200	1.4170	0.0426	-0.4297	0.6885	0.4723	-1.1706	-0.9747	-2.6767	-2.7110

Framework versions

• Transformers 4.43.1
• Pytorch 2.1.2+cu121
• Datasets 2.18.0
• Tokenizers 0.19.1