AA_text_image_to_text
This model is a fine-tuned version of llava-hf/llava-v1.6-mistral-7b-hf on the AA_text_image_to_text dataset. It achieves the following results on the evaluation set:
- Loss: 0.4500
- Rewards/chosen: -0.6971
- Rewards/rejected: -4.4006
- Rewards/accuracies: 0.8206
- Rewards/margins: 3.7035
- Logps/rejected: -242.2139
- Logps/chosen: -207.2900
- Logits/rejected: -1.9132
- Logits/chosen: -1.9735
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 |
---|---|---|---|---|---|---|---|---|---|---|---|
0.4516 | 0.3623 | 50 | 0.4421 | 0.8113 | -0.8305 | 0.7903 | 1.6418 | -206.5130 | -192.2066 | -1.6310 | -1.7066 |
0.3759 | 0.7246 | 100 | 0.4121 | -0.0472 | -2.2762 | 0.8145 | 2.2290 | -220.9696 | -200.7911 | -1.7930 | -1.8524 |
0.149 | 1.0870 | 150 | 0.4205 | 0.5835 | -1.8816 | 0.8206 | 2.4651 | -217.0244 | -194.4847 | -1.6746 | -1.7425 |
0.1474 | 1.4493 | 200 | 0.4274 | -0.5411 | -3.7374 | 0.8306 | 3.1963 | -235.5818 | -205.7306 | -1.7947 | -1.8599 |
0.1268 | 1.8116 | 250 | 0.4333 | -0.0670 | -3.3107 | 0.8206 | 3.2437 | -231.3154 | -200.9896 | -2.0993 | -2.1450 |
0.064 | 2.1739 | 300 | 0.4332 | -0.5167 | -4.0958 | 0.8306 | 3.5792 | -239.1665 | -205.4860 | -1.9327 | -1.9909 |
0.056 | 2.5362 | 350 | 0.4481 | -0.5224 | -4.1134 | 0.8185 | 3.5910 | -239.3422 | -205.5439 | -1.9163 | -1.9756 |
0.0721 | 2.8986 | 400 | 0.4507 | -0.7023 | -4.4082 | 0.8185 | 3.7059 | -242.2901 | -207.3426 | -1.9129 | -1.9731 |
Framework versions
- Transformers 4.45.2
- Pytorch 2.4.0+cu121
- Datasets 2.21.0
- Tokenizers 0.20.3
- Downloads last month
- 3
Inference API (serverless) does not yet support transformers models for this pipeline type.
Model tree for htlou/mm-interp-AA_text_image_to_text
Base model
llava-hf/llava-v1.6-mistral-7b-hf