This model is a fine-tuned version of llava-hf/llava-v1.6-mistral-7b-hf on the AA_preference_l0_new_step10_0_30 dataset. It achieves the following results on the evaluation set:

Loss: 0.6179
Rewards/chosen: 2.2275
Rewards/rejected: -0.1215
Rewards/accuracies: 0.7778
Rewards/margins: 2.3490
Logps/rejected: -258.3642
Logps/chosen: -268.2225
Logits/rejected: -2.1741
Logits/chosen: -2.2030

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
0.258	1.2422	50	0.6213	2.3820	0.3959	0.7847	1.9861	-253.1901	-266.6775	-2.3213	-2.3355
0.1741	2.4845	100	0.6181	2.2334	-0.1129	0.7778	2.3463	-258.2783	-268.1634	-2.1729	-2.2018

Framework versions

Transformers 4.45.2
Pytorch 2.4.0+cu121
Datasets 2.21.0
Tokenizers 0.20.3

htlou/backup_0202_llamafactory_AA_preference_l0_new_step10_0_30-llava-mistral

AA_preference_l0_new_step10_0_30

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

htlou/
backup_0202_llamafactory_AA_preference_l0_new_step10_0_30-llava-mistral