This model is a fine-tuned version of llava-hf/llava-v1.6-mistral-7b-hf on the AA_preference_cosi_new_step10_0_10 dataset.
It achieves the following results on the evaluation set:
Loss: 0.6302
Rewards/chosen: 1.4231
Rewards/rejected: -0.2936
Rewards/accuracies: 0.6667
Rewards/margins: 1.7167
Logps/rejected: -239.9530
Logps/chosen: -271.4290
Logits/rejected: -2.4200
Logits/chosen: -2.3885
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: 1e-06
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
