metadata
license: apache-2.0
base_model: mistralai/Mistral-7B-v0.1
tags:
- generated_from_trainer
model-index:
- name: sambar-7b-dpo-lora
results: []
sambar-7b-dpo-lora
This model is a fine-tuned version of mistralai/Mistral-7B-v0.1 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.5747
- Rewards/chosen: -0.0141
- Rewards/rejected: -0.4147
- Rewards/accuracies: 0.7060
- Rewards/margins: 0.4006
- Logps/rejected: -221.3069
- Logps/chosen: -263.0773
- Logits/rejected: -2.1478
- Logits/chosen: -2.2594
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: 5e-07
- train_batch_size: 2
- eval_batch_size: 4
- seed: 42
- distributed_type: multi-GPU
- num_devices: 4
- gradient_accumulation_steps: 32
- total_train_batch_size: 256
- total_eval_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 3
Training results
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.6213 | 1.0 | 242 | 0.6182 | 0.0426 | -0.1569 | 0.6860 | 0.1995 | -218.7293 | -262.5110 | -2.1605 | -2.2727 |
| 0.5903 | 2.0 | 484 | 0.5826 | 0.0046 | -0.3500 | 0.6940 | 0.3546 | -220.6603 | -262.8906 | -2.1517 | -2.2634 |
| 0.5743 | 3.0 | 726 | 0.5747 | -0.0141 | -0.4147 | 0.7060 | 0.4006 | -221.3069 | -263.0773 | -2.1478 | -2.2594 |
Framework versions
- Transformers 4.35.0
- Pytorch 2.1.2+cu121
- Datasets 2.14.6
- Tokenizers 0.14.1