AA_preference_cosi_new_step10_0_30
This model is a fine-tuned version of llava-hf/llava-v1.6-mistral-7b-hf on the AA_preference_cosi_new_step10_0_30 dataset. It achieves the following results on the evaluation set:
- Loss: 0.5883
- Rewards/chosen: 1.9964
- Rewards/rejected: -0.3785
- Rewards/accuracies: 0.8403
- Rewards/margins: 2.3750
- Logps/rejected: -233.2932
- Logps/chosen: -249.7375
- Logits/rejected: -2.2233
- Logits/chosen: -2.2292
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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.2594 | 1.2422 | 50 | 0.5902 | 1.6739 | -0.2014 | 0.7986 | 1.8753 | -231.5221 | -252.9634 | -2.2944 | -2.2956 |
| 0.1658 | 2.4845 | 100 | 0.5874 | 2.0247 | -0.3441 | 0.8333 | 2.3688 | -232.9490 | -249.4547 | -2.2372 | -2.2420 |
Framework versions
- Transformers 4.45.2
- Pytorch 2.4.0+cu121
- Datasets 2.21.0
- Tokenizers 0.20.3
- Downloads last month
- 9
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_30
Base model
llava-hf/llava-v1.6-mistral-7b-hf