rlced_conifer_zephyr-7b-dpo-full

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

• Loss: 0.2006
• Rewards/chosen: -4.0000
• Rewards/rejected: -11.3920
• Rewards/accuracies: 0.8687
• Rewards/margins: 7.3920
• Logps/rejected: -1507.9508
• Logps/chosen: -760.1741
• Logits/rejected: -1.6795
• Logits/chosen: -2.0969

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: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 4
• total_train_batch_size: 256
• 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: 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
0.2558	0.2107	100	0.2515	-3.3402	-7.6336	0.8363	4.2934	-1132.1079	-694.1878	-3.1129	-3.1245
0.2204	0.4215	200	0.2260	-3.7625	-9.3814	0.8587	5.6190	-1306.8950	-736.4189	-2.6948	-2.7329
0.204	0.6322	300	0.2096	-3.3959	-9.7494	0.8650	6.3535	-1343.6858	-699.7554	-2.1723	-2.4109
0.1992	0.8430	400	0.2007	-4.2176	-11.7584	0.8650	7.5408	-1544.5891	-781.9304	-1.6459	-2.0690

Framework versions

• Transformers 4.44.1
• Pytorch 2.1.2+cu121
• Datasets 2.21.0
• Tokenizers 0.19.1