This model is a fine-tuned version of alignment-handbook/zephyr-7b-sft-full on the data/uf and the data/rlced_conifer datasets.
It achieves the following results on the evaluation set:
Loss: 0.3443
Rewards/chosen: -1.6459
Rewards/rejected: -3.1222
Rewards/accuracies: 0.8701
Rewards/margins: 1.4764
Logps/rejected: -758.1630
Logps/chosen: -587.9950
Logits/rejected: -2.7637
Logits/chosen: -2.7840
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
