AA_preference_cosi_new_step10_0_20
This model is a fine-tuned version of llava-hf/llava-v1.6-mistral-7b-hf on the AA_preference_cosi_new_step10_0_20 dataset. It achieves the following results on the evaluation set:
- Loss: 0.5770
- Rewards/chosen: 1.9856
- Rewards/rejected: -0.3209
- Rewards/accuracies: 0.7812
- Rewards/margins: 2.3065
- Logps/rejected: -251.8845
- Logps/chosen: -272.1190
- Logits/rejected: -2.5221
- Logits/chosen: -2.5211
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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.2858 | 1.8692 | 50 | 0.5790 | 1.9154 | -0.4284 | 0.7812 | 2.3437 | -252.9591 | -272.8219 | -2.5505 | -2.5497 |
Framework versions
- Transformers 4.45.2
- Pytorch 2.4.0+cu121
- Datasets 2.21.0
- Tokenizers 0.20.3
- Downloads last month
- 13
Inference API (serverless) does not yet support transformers models for this pipeline type.
Model tree for htlou/mm-interp-AA_preference_cosi_new_step10_0_20
Base model
llava-hf/llava-v1.6-mistral-7b-hf