AA_preference_cosi_new_step10_0_40
This model is a fine-tuned version of llava-hf/llava-v1.6-mistral-7b-hf on the AA_preference_cosi_new_step10_0_40 dataset. It achieves the following results on the evaluation set:
- Loss: 0.5665
- Rewards/chosen: 1.7347
- Rewards/rejected: -0.8858
- Rewards/accuracies: 0.7969
- Rewards/margins: 2.6205
- Logps/rejected: -230.7836
- Logps/chosen: -258.4571
- Logits/rejected: -2.2696
- Logits/chosen: -2.2821
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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.5111 | 0.9346 | 50 | 0.5646 | 1.4841 | -0.0695 | 0.7448 | 1.5536 | -222.6201 | -260.9629 | -2.4587 | -2.4679 |
| 0.2954 | 1.8692 | 100 | 0.5597 | 1.7109 | -0.9040 | 0.7917 | 2.6149 | -230.9650 | -258.6947 | -2.0402 | -2.0607 |
| 0.1503 | 2.8037 | 150 | 0.5670 | 1.7345 | -0.8860 | 0.7969 | 2.6204 | -230.7849 | -258.4590 | -2.2699 | -2.2824 |
Framework versions
- Transformers 4.45.2
- Pytorch 2.4.0+cu121
- Datasets 2.21.0
- Tokenizers 0.20.3
- Downloads last month
- 12
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_40
Base model
llava-hf/llava-v1.6-mistral-7b-hf