zephyr-7b-dpo-qlora-fsdp

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

• Loss: 0.6843
• Rewards/chosen: 0.0234
• Rewards/rejected: 0.0034
• Rewards/accuracies: 0.6211
• Rewards/margins: 0.0199
• Logps/rejected: -260.8430
• Logps/chosen: -258.9067
• Logits/rejected: -2.4164
• Logits/chosen: -2.4494

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-06
• train_batch_size: 15
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 4
• total_train_batch_size: 480
• total_eval_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 0.1

Training results

Framework versions

• PEFT 0.9.0
• Transformers 4.38.1
• Pytorch 2.2.0+cu118
• Datasets 2.17.1
• Tokenizers 0.15.2