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: 0.5061
Rewards/chosen: -1.0205
Rewards/rejected: -1.9944
Rewards/accuracies: 0.7812
Rewards/margins: 0.9740
Logps/rejected: -462.1064
Logps/chosen: -364.6778
Logits/rejected: 2.4089
Logits/chosen: 1.6213

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: 2
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: 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
0.5644	0.2092	100	0.5695	-0.6773	-1.2063	0.7188	0.5290	-383.2874	-330.3586	-0.4241	-0.6326
0.5445	0.4184	200	0.5354	-0.6109	-1.3936	0.7695	0.7827	-402.0177	-323.7153	0.9697	0.1924
0.4909	0.6276	300	0.5138	-0.9252	-1.8338	0.7578	0.9086	-446.0438	-355.1510	2.1667	1.4140
0.5036	0.8368	400	0.5066	-0.9998	-1.9548	0.7812	0.9549	-458.1404	-362.6149	2.3428	1.5651

Framework versions

Transformers 4.44.0
Pytorch 2.1.2+cu121
Datasets 2.21.0
Tokenizers 0.19.1

zodi1121/zephyr-7b-dpo-full

zephyr-7b-dpo-full

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

zodi1121/
zephyr-7b-dpo-full