File size: 3,334 Bytes
bf9d77d eb7e456 bf9d77d df68f81 bf9d77d 2de7880 bf9d77d f41965d df68f81 bf9d77d 62106cc bf9d77d 62106cc |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 |
---
tags:
- trl
- dpo
- generated_from_trainer
model-index:
- name: zephyr-7b-dpo-full
results: []
---
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# zephyr-7b-dpo-full
This model was trained from scratch on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 0.2027
- Rewards/chosen: 0.6729
- Rewards/rejected: -2.3580
- Rewards/accuracies: 0.9141
- Rewards/margins: 3.0309
- Logps/rejected: -380.2658
- Logps/chosen: -322.0539
- Logits/rejected: -1.9204
- Logits/chosen: -1.9591
## 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: 8
- eval_batch_size: 8
- seed: 42
- distributed_type: multi-GPU
- num_devices: 8
- gradient_accumulation_steps: 2
- total_train_batch_size: 128
- total_eval_batch_size: 64
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 1
### Training results
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
| 0.4898 | 0.12 | 100 | 0.5505 | -0.1967 | -1.0051 | 0.6875 | 0.8085 | -353.2088 | -339.4445 | -1.7659 | -1.8469 |
| 0.4277 | 0.23 | 200 | 0.4655 | -0.4834 | -1.8836 | 0.7383 | 1.4002 | -370.7788 | -345.1795 | -1.7248 | -1.8009 |
| 0.4188 | 0.35 | 300 | 0.3922 | -0.0720 | -2.0263 | 0.7969 | 1.9544 | -373.6328 | -336.9513 | -1.6143 | -1.6899 |
| 0.3506 | 0.46 | 400 | 0.3457 | 0.2171 | -2.0472 | 0.8203 | 2.2643 | -374.0495 | -331.1692 | -1.9794 | -2.0296 |
| 0.3611 | 0.58 | 500 | 0.2959 | 0.2498 | -2.4347 | 0.8516 | 2.6844 | -381.7997 | -330.5164 | -1.8183 | -1.8592 |
| 0.3562 | 0.69 | 600 | 0.2513 | 0.3868 | -2.4732 | 0.8711 | 2.8600 | -382.5696 | -327.7753 | -1.9217 | -1.9736 |
| 0.3624 | 0.81 | 700 | 0.2194 | 0.6454 | -2.3556 | 0.9062 | 3.0010 | -380.2178 | -322.6031 | -1.9301 | -1.9717 |
| 0.4069 | 0.92 | 800 | 0.2027 | 0.6729 | -2.3580 | 0.9141 | 3.0309 | -380.2658 | -322.0539 | -1.9204 | -1.9591 |
### Framework versions
- Transformers 4.38.2
- Pytorch 2.1.2+cu118
- Datasets 2.16.1
- Tokenizers 0.15.2
|