metadata
license: apache-2.0
base_model: alignment-handbook/zephyr-7b-sft-full
tags:
- generated_from_trainer
model-index:
- name: spin-v-high-loss
results: []
spin-v-high-loss
This model is a fine-tuned version of alignment-handbook/zephyr-7b-sft-full on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.0069
- Rewards/real: -10.1415
- Rewards/generated: -55.1541
- Rewards/accuracies: 1.0
- Rewards/margins: 45.0126
- Logps/generated: -5640.6729
- Logps/real: -1151.2217
- Logits/generated: 3.0744
- Logits/real: 1.9177
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: 4
- total_train_batch_size: 32
- total_eval_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 1
Training results
| Training Loss | Epoch | Step | Validation Loss | Rewards/real | Rewards/generated | Rewards/accuracies | Rewards/margins | Logps/generated | Logps/real | Logits/generated | Logits/real |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.0717 | 0.13 | 50 | 0.0490 | -3.1258 | -37.3431 | 0.9907 | 34.2173 | -3859.5708 | -449.6532 | 3.4831 | 1.9303 |
| 0.0323 | 0.27 | 100 | 0.0300 | -3.9959 | -38.8380 | 0.9973 | 34.8421 | -4009.0552 | -536.6592 | -0.0155 | -0.1626 |
| 0.026 | 0.4 | 150 | 0.0158 | -8.2107 | -50.0493 | 0.9947 | 41.8386 | -5130.1880 | -958.1443 | 1.0207 | 1.0071 |
| 0.0106 | 0.53 | 200 | 0.0087 | -9.2505 | -61.7325 | 0.9960 | 52.4820 | -6298.5093 | -1062.1265 | 2.2349 | 1.2992 |
| 0.0071 | 0.67 | 250 | 0.0106 | -11.4051 | -49.3118 | 0.9987 | 37.9067 | -5056.4409 | -1277.5874 | 2.8798 | 3.2925 |
| 0.0121 | 0.8 | 300 | 0.0074 | -9.0224 | -49.1152 | 1.0 | 40.0928 | -5036.7827 | -1039.3110 | 2.8713 | 2.6792 |
| 0.0013 | 0.93 | 350 | 0.0069 | -10.1415 | -55.1541 | 1.0 | 45.0126 | -5640.6729 | -1151.2217 | 3.0744 | 1.9177 |
Framework versions
- Transformers 4.37.0
- Pytorch 2.1.2+cu121
- Datasets 2.14.6
- Tokenizers 0.15.2