zephyr0-7b-ultra-p-0.07

This model is a fine-tuned version of HuggingFaceH4/mistral-7b-sft-beta on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.4934
• Rewards/chosen: -1.1822
• Rewards/rejected: -2.0719
• Rewards/accuracies: 0.75
• Rewards/margins: 0.8897
• Logps/rejected: -434.9780
• Logps/chosen: -340.5889
• Logits/rejected: -1.9202
• Logits/chosen: -2.0197

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-07
• train_batch_size: 2
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 8
• total_train_batch_size: 128
• total_eval_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 1.0

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.5521	0.2060	100	0.5258	-0.8611	-1.4030	0.6953	0.5419	-368.0812	-308.4762	-2.4729	-2.5388
0.5086	0.4119	200	0.5115	-1.1790	-1.8930	0.75	0.7140	-417.0868	-340.2655	-2.1594	-2.2382
0.49	0.6179	300	0.5050	-1.0454	-1.8769	0.7422	0.8316	-415.4766	-326.9024	-2.0191	-2.1080
0.4855	0.8239	400	0.4962	-1.1394	-2.0028	0.75	0.8634	-428.0682	-336.3091	-1.9714	-2.0677

Framework versions

• Transformers 4.45.1
• Pytorch 2.4.1+cu121
• Datasets 3.0.0
• Tokenizers 0.20.0