OpenELM-1_1B-DPO-full-max-4-reward
This model was trained from scratch on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 1.6190
- Rewards/chosen: -13.625
- Rewards/rejected: -15.0625
- Rewards/accuracies: 0.5996
- Rewards/margins: 1.4688
- Logps/rejected: -1800.0
- Logps/chosen: -1680.0
- Logits/rejected: 1.0625
- Logits/chosen: -0.2695
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-05
- train_batch_size: 8
- eval_batch_size: 16
- seed: 42
- distributed_type: multi-GPU
- num_devices: 4
- gradient_accumulation_steps: 2
- total_train_batch_size: 64
- 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: 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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.6378 | 0.0838 | 80 | 0.6868 | -0.6758 | -0.7656 | 0.5684 | 0.0918 | -366.0 | -386.0 | -9.875 | -10.125 |
| 0.6219 | 0.1675 | 160 | 0.6949 | -0.9102 | -1.0547 | 0.5977 | 0.1406 | -394.0 | -410.0 | -10.125 | -10.5 |
| 0.6151 | 0.2513 | 240 | 0.7637 | -2.4531 | -2.6562 | 0.5566 | 0.2031 | -552.0 | -564.0 | -10.9375 | -11.25 |
| 0.6607 | 0.3351 | 320 | 0.7307 | -2.7344 | -2.9375 | 0.5742 | 0.1992 | -584.0 | -592.0 | -14.25 | -14.4375 |
| 0.6304 | 0.4188 | 400 | 0.7129 | -2.7344 | -3.0156 | 0.5898 | 0.2715 | -588.0 | -592.0 | -12.5 | -13.0 |
| 0.623 | 0.5026 | 480 | 0.7718 | -2.5469 | -2.9375 | 0.5859 | 0.3887 | -584.0 | -572.0 | -8.0625 | -9.0 |
| 0.6091 | 0.5864 | 560 | 0.7543 | -3.3281 | -3.6562 | 0.5957 | 0.3320 | -656.0 | -652.0 | -12.0 | -12.75 |
| 0.583 | 0.6702 | 640 | 0.7081 | -3.25 | -3.7031 | 0.6406 | 0.4648 | -660.0 | -644.0 | -9.0 | -10.0625 |
| 0.6183 | 0.7539 | 720 | 0.7397 | -3.7812 | -4.0938 | 0.5996 | 0.3242 | -700.0 | -696.0 | -8.5625 | -9.4375 |
| 0.5988 | 0.8377 | 800 | 0.7986 | -4.4688 | -4.9375 | 0.5898 | 0.4609 | -784.0 | -764.0 | -7.9062 | -8.9375 |
| 0.5882 | 0.9215 | 880 | 0.7997 | -3.2656 | -3.6562 | 0.5879 | 0.3906 | -656.0 | -644.0 | -8.3125 | -9.1875 |
| 0.4256 | 1.0052 | 960 | 0.7816 | -4.5312 | -5.1875 | 0.6172 | 0.6367 | -808.0 | -772.0 | -6.75 | -7.9062 |
| 0.2006 | 1.0890 | 1040 | 0.9734 | -5.9688 | -6.6875 | 0.6094 | 0.7383 | -960.0 | -916.0 | -4.7812 | -6.0625 |
| 0.1977 | 1.1728 | 1120 | 0.9420 | -6.25 | -7.0 | 0.6094 | 0.7578 | -988.0 | -944.0 | -5.0 | -6.25 |
| 0.1717 | 1.2565 | 1200 | 1.0548 | -7.4688 | -8.25 | 0.5918 | 0.7852 | -1112.0 | -1064.0 | -4.5 | -5.8125 |
| 0.1881 | 1.3403 | 1280 | 0.9567 | -6.9688 | -7.8125 | 0.6035 | 0.8672 | -1072.0 | -1012.0 | -3.2188 | -4.4688 |
| 0.1897 | 1.4241 | 1360 | 0.9563 | -6.9688 | -7.8438 | 0.6055 | 0.8867 | -1072.0 | -1016.0 | -4.2812 | -5.6875 |
| 0.1383 | 1.5079 | 1440 | 1.1196 | -8.5625 | -9.5 | 0.6055 | 0.9922 | -1240.0 | -1176.0 | -2.5938 | -3.9062 |
| 0.146 | 1.5916 | 1520 | 1.0767 | -9.5 | -10.5 | 0.6055 | 1.0078 | -1336.0 | -1264.0 | -1.6797 | -3.0312 |
| 0.1831 | 1.6754 | 1600 | 0.9776 | -8.0625 | -8.9375 | 0.6055 | 0.8516 | -1184.0 | -1128.0 | -2.2344 | -3.5938 |
| 0.1667 | 1.7592 | 1680 | 1.0210 | -7.75 | -8.625 | 0.5957 | 0.9023 | -1152.0 | -1088.0 | -1.7344 | -3.2344 |
| 0.1514 | 1.8429 | 1760 | 1.0214 | -8.6875 | -9.6875 | 0.6133 | 0.9805 | -1256.0 | -1184.0 | -1.1719 | -2.5312 |
| 0.1594 | 1.9267 | 1840 | 1.0633 | -8.8125 | -9.75 | 0.5977 | 0.9727 | -1264.0 | -1200.0 | -1.2344 | -2.625 |
| 0.0307 | 2.0105 | 1920 | 1.0948 | -8.75 | -9.75 | 0.6172 | 1.0312 | -1264.0 | -1192.0 | -1.4531 | -2.9844 |
| 0.0214 | 2.0942 | 2000 | 1.5354 | -12.25 | -13.3125 | 0.6094 | 1.1016 | -1624.0 | -1544.0 | 0.1973 | -1.2031 |
| 0.0186 | 2.1780 | 2080 | 1.5790 | -13.5625 | -14.9375 | 0.6055 | 1.3906 | -1784.0 | -1680.0 | 0.4902 | -0.9102 |
| 0.0395 | 2.2618 | 2160 | 1.5234 | -12.0625 | -13.1875 | 0.6035 | 1.1406 | -1608.0 | -1520.0 | 0.5391 | -0.7656 |
| 0.0217 | 2.3455 | 2240 | 1.5867 | -13.1875 | -14.5625 | 0.6035 | 1.375 | -1744.0 | -1632.0 | 0.8945 | -0.4141 |
| 0.0268 | 2.4293 | 2320 | 1.5888 | -13.0 | -14.375 | 0.6035 | 1.4219 | -1728.0 | -1616.0 | 0.6797 | -0.6758 |
| 0.0238 | 2.5131 | 2400 | 1.6647 | -13.625 | -15.0625 | 0.6055 | 1.4453 | -1792.0 | -1680.0 | 0.9648 | -0.3633 |
| 0.0227 | 2.5969 | 2480 | 1.5873 | -13.125 | -14.5625 | 0.6094 | 1.4375 | -1744.0 | -1632.0 | 0.9258 | -0.4199 |
| 0.0233 | 2.6806 | 2560 | 1.5836 | -13.1875 | -14.625 | 0.6035 | 1.4297 | -1752.0 | -1640.0 | 0.9297 | -0.4180 |
| 0.021 | 2.7644 | 2640 | 1.5917 | -13.4375 | -14.9375 | 0.6094 | 1.4609 | -1776.0 | -1664.0 | 1.0078 | -0.3223 |
| 0.0221 | 2.8482 | 2720 | 1.6077 | -13.5625 | -15.0 | 0.6035 | 1.4609 | -1792.0 | -1672.0 | 1.0469 | -0.2793 |
| 0.0182 | 2.9319 | 2800 | 1.6190 | -13.625 | -15.0625 | 0.5996 | 1.4688 | -1800.0 | -1680.0 | 1.0625 | -0.2695 |
Framework versions
- Transformers 4.45.1
- Pytorch 2.3.0
- Datasets 3.0.1
- Tokenizers 0.20.0
- Downloads last month
- 12
Inference API (serverless) does not yet support model repos that contain custom code.