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: 0.6357
Rewards/chosen: -0.0757
Rewards/rejected: -0.3058
Rewards/accuracies: 0.3353
Rewards/margins: 0.2301
Logps/rejected: -89.7913
Logps/chosen: -76.6356
Logits/rejected: -2.3906
Logits/chosen: -2.4061
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: 8.639811904061041e-07
train_batch_size: 8
eval_batch_size: 8
seed: 42
distributed_type: multi-GPU
num_devices: 4
gradient_accumulation_steps: 2
total_train_batch_size: 64
total_eval_batch_size: 32
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
