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

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: 1.3549
• Rewards/chosen: 0.0671
• Rewards/rejected: -0.2877
• Rewards/accuracies: 0.7480
• Rewards/margins: 0.3547
• Logps/rejected: -1.1564
• Logps/chosen: -0.9723
• Logits/rejected: -2.6030
• Logits/chosen: -2.6556

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-07
• 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
1.3576	0.4187	100	1.3627	0.1418	-0.1881	0.7381	0.3299	-1.1465	-0.9648	-2.6131	-2.6646
1.348	0.8375	200	1.3558	0.0514	-0.3071	0.7460	0.3586	-1.1584	-0.9738	-2.6024	-2.6552

Framework versions

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