This model is a fine-tuned version of llava-hf/llava-v1.6-mistral-7b-hf on the AA_preference_cocour_new_step10_0_90 dataset. It achieves the following results on the evaluation set:

Loss: 0.4735
Rewards/chosen: 0.5420
Rewards/rejected: -2.3019
Rewards/accuracies: 0.8218
Rewards/margins: 2.8439
Logps/rejected: -225.0065
Logps/chosen: -243.1698
Logits/rejected: -2.6339
Logits/chosen: -2.6492

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
0.6087	0.4158	50	0.5778	0.7419	-0.5074	0.7870	1.2493	-207.0612	-241.1710	-2.5285	-2.5456
0.5039	0.8316	100	0.5226	0.2943	-1.5654	0.8056	1.8596	-217.6406	-245.6472	-2.5932	-2.6105
0.2346	1.2474	150	0.4851	0.5179	-1.8870	0.8356	2.4048	-220.8566	-243.4111	-2.5511	-2.5729
0.26	1.6632	200	0.4692	0.8149	-1.7120	0.8264	2.5269	-219.1066	-240.4409	-2.6651	-2.6766
0.1628	2.0790	250	0.4654	0.2522	-2.4566	0.8264	2.7088	-226.5530	-246.0683	-2.6802	-2.6929
0.1808	2.4948	300	0.4721	0.8229	-2.0114	0.8241	2.8342	-222.1007	-240.3612	-2.6392	-2.6528
0.1514	2.9106	350	0.4736	0.5409	-2.3033	0.8218	2.8442	-225.0204	-243.1809	-2.6346	-2.6500

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_cocour_new_step10_0_90-llava-mistral

AA_preference_cocour_new_step10_0_90

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

htlou/
backup_0202_llamafactory_AA_preference_cocour_new_step10_0_90-llava-mistral