htlou/mm-interp-AA_preference_cocour_new_step10_0_30

AA_preference_cocour_new_step10_0_30

This model is a fine-tuned version of llava-hf/llava-v1.6-mistral-7b-hf on the AA_preference_cocour_new_step10_0_30 dataset. It achieves the following results on the evaluation set:

• Loss: 0.5532
• Rewards/chosen: 1.6544
• Rewards/rejected: -1.2534
• Rewards/accuracies: 0.8056
• Rewards/margins: 2.9078
• Logps/rejected: -252.7311
• Logps/chosen: -241.2080
• Logits/rejected: -2.0966
• Logits/chosen: -2.1262

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
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 10
• num_epochs: 3.0

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.2648	1.2422	50	0.5693	2.0296	-0.4936	0.8125	2.5232	-245.1335	-237.4563	-2.3209	-2.3384
0.1487	2.4845	100	0.5524	1.6519	-1.2519	0.8125	2.9039	-252.7170	-241.2330	-2.0940	-2.1241

Framework versions

• Transformers 4.45.2
• Pytorch 2.4.0+cu121
• Datasets 2.21.0
• Tokenizers 0.20.3