This model is a fine-tuned version of allenai/tulu-2-7b on the data/tulu-2-7b-uf-rlced-conifer-ref dataset.
It achieves the following results on the evaluation set:
Loss: 0.3478
Rewards/chosen: -2.6376
Rewards/rejected: -4.8787
Rewards/accuracies: 0.8355
Rewards/margins: 2.2410
Logps/rejected: -973.1326
Logps/chosen: -725.1460
Logits/rejected: 0.0005
Logits/chosen: -0.1672
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
