metadata
tags:
- trl
- dpo
- generated_from_trainer
model-index:
- name: zephyr-7b-dpo-full
results: []
zephyr-7b-dpo-full
This model was trained from scratch on the None dataset. It achieves the following results on the evaluation set:
- Loss: 1.8979
- Rewards/chosen: -6.9869
- Rewards/rejected: -8.4701
- Rewards/accuracies: 0.6094
- Rewards/margins: 1.4832
- Logps/rejected: -1164.5387
- Logps/chosen: -1010.4669
- Logits/rejected: -0.5643
- Logits/chosen: -0.7199
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 |
---|---|---|---|---|---|---|---|---|---|---|---|
0.2555 | 0.1 | 100 | 1.4172 | -4.8884 | -5.6701 | 0.5898 | 0.7817 | -884.5335 | -800.6121 | -1.3358 | -1.3942 |
0.1854 | 0.21 | 200 | 1.6754 | -6.1508 | -7.3259 | 0.6211 | 1.1752 | -1050.1200 | -926.8517 | -1.1088 | -1.1853 |
0.1799 | 0.31 | 300 | 1.5590 | -5.9157 | -6.9794 | 0.5977 | 1.0637 | -1015.4615 | -903.3419 | -1.0193 | -1.1110 |
0.1679 | 0.42 | 400 | 2.1030 | -7.8503 | -9.2060 | 0.6094 | 1.3557 | -1238.1252 | -1096.8108 | -0.5753 | -0.7096 |
0.1693 | 0.52 | 500 | 1.6563 | -6.3408 | -7.6718 | 0.625 | 1.3310 | -1084.7078 | -945.8611 | -0.8598 | -0.9873 |
0.1609 | 0.63 | 600 | 1.6818 | -6.4795 | -7.7992 | 0.6211 | 1.3198 | -1097.4480 | -959.7227 | -0.4515 | -0.6164 |
0.1559 | 0.73 | 700 | 1.9278 | -7.3485 | -8.7955 | 0.6133 | 1.4470 | -1197.0731 | -1046.6217 | -0.4166 | -0.5852 |
0.1433 | 0.84 | 800 | 1.9050 | -7.1496 | -8.6252 | 0.6172 | 1.4756 | -1180.0403 | -1026.7318 | -0.5141 | -0.6745 |
0.1479 | 0.94 | 900 | 1.8979 | -6.9869 | -8.4701 | 0.6094 | 1.4832 | -1164.5387 | -1010.4669 | -0.5643 | -0.7199 |
Framework versions
- Transformers 4.38.2
- Pytorch 2.1.2+cu118
- Datasets 2.16.1
- Tokenizers 0.15.2