metadata
library_name: transformers
tags:
- trl
- dpo
- generated_from_trainer
model-index:
- name: OpenELM-1_1B-DPO-full-max-6-reward
results: []
OpenELM-1_1B-DPO-full-max-6-reward
This model was trained from scratch on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 1.6897
- Rewards/chosen: -15.0625
- Rewards/rejected: -16.625
- Rewards/accuracies: 0.5918
- Rewards/margins: 1.6172
- Logps/rejected: -1952.0
- Logps/chosen: -1824.0
- Logits/rejected: 2.4844
- Logits/chosen: 0.6445
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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.5942 | 0.1047 | 100 | 0.6780 | -1.3594 | -1.5625 | 0.6055 | 0.2051 | -446.0 | -454.0 | -11.125 | -11.4375 |
| 0.561 | 0.2094 | 200 | 0.6944 | -2.2188 | -2.5156 | 0.6055 | 0.2949 | -540.0 | -540.0 | -9.75 | -10.25 |
| 0.5506 | 0.3141 | 300 | 0.7181 | -4.4375 | -4.8438 | 0.6074 | 0.4004 | -772.0 | -764.0 | -7.5 | -8.3125 |
| 0.5233 | 0.4188 | 400 | 0.8159 | -4.0312 | -4.5 | 0.5840 | 0.4668 | -736.0 | -720.0 | -12.875 | -13.625 |
| 0.5164 | 0.5236 | 500 | 0.7724 | -3.1562 | -3.8281 | 0.6270 | 0.6602 | -672.0 | -636.0 | -6.6562 | -8.3125 |
| 0.5142 | 0.6283 | 600 | 0.7633 | -5.3438 | -5.7188 | 0.5645 | 0.3770 | -860.0 | -852.0 | -6.4375 | -7.4062 |
| 0.4846 | 0.7330 | 700 | 0.8138 | -5.6875 | -6.3438 | 0.5996 | 0.6641 | -924.0 | -888.0 | -7.25 | -8.625 |
| 0.475 | 0.8377 | 800 | 0.7354 | -5.5 | -6.0938 | 0.6133 | 0.6016 | -900.0 | -868.0 | -8.9375 | -10.0625 |
| 0.4734 | 0.9424 | 900 | 0.8621 | -7.5625 | -8.3125 | 0.6074 | 0.7305 | -1120.0 | -1072.0 | -5.75 | -7.3438 |
| 0.146 | 1.0471 | 1000 | 1.1525 | -10.6875 | -11.5 | 0.5684 | 0.8477 | -1440.0 | -1384.0 | -2.7031 | -4.4062 |
| 0.1356 | 1.1518 | 1100 | 0.9972 | -9.1875 | -10.0625 | 0.6094 | 0.9062 | -1296.0 | -1232.0 | -3.6875 | -5.5312 |
| 0.1237 | 1.2565 | 1200 | 0.9993 | -8.5 | -9.25 | 0.6016 | 0.7617 | -1216.0 | -1168.0 | -6.3438 | -8.125 |
| 0.1387 | 1.3613 | 1300 | 1.1685 | -10.875 | -11.9375 | 0.5977 | 1.0625 | -1488.0 | -1408.0 | -2.4219 | -4.5625 |
| 0.1298 | 1.4660 | 1400 | 1.0590 | -10.4375 | -11.5 | 0.5996 | 1.0312 | -1440.0 | -1368.0 | -2.9844 | -4.9375 |
| 0.1284 | 1.5707 | 1500 | 1.1094 | -11.125 | -12.25 | 0.6191 | 1.1172 | -1512.0 | -1432.0 | -1.9766 | -3.9844 |
| 0.1084 | 1.6754 | 1600 | 1.1327 | -10.8125 | -11.9375 | 0.6230 | 1.1328 | -1480.0 | -1400.0 | -2.2812 | -4.1875 |
| 0.1028 | 1.7801 | 1700 | 1.1308 | -11.0 | -12.25 | 0.6230 | 1.2344 | -1512.0 | -1424.0 | -1.7031 | -3.625 |
| 0.1227 | 1.8848 | 1800 | 1.0562 | -9.5 | -10.5 | 0.6113 | 0.9961 | -1336.0 | -1264.0 | -2.5156 | -4.375 |
| 0.1119 | 1.9895 | 1900 | 1.1633 | -11.3125 | -12.5625 | 0.5996 | 1.2344 | -1544.0 | -1448.0 | -1.2734 | -3.1094 |
| 0.0166 | 2.0942 | 2000 | 1.5515 | -14.25 | -15.75 | 0.5840 | 1.5234 | -1864.0 | -1744.0 | 1.9375 | 0.0549 |
| 0.0122 | 2.1990 | 2100 | 1.6694 | -14.8125 | -16.5 | 0.6094 | 1.6484 | -1936.0 | -1800.0 | 1.5703 | -0.4121 |
| 0.0158 | 2.3037 | 2200 | 1.7912 | -15.8125 | -17.5 | 0.6035 | 1.6875 | -2040.0 | -1904.0 | 2.5156 | 0.6445 |
| 0.0165 | 2.4084 | 2300 | 1.7817 | -15.8125 | -17.375 | 0.5938 | 1.5703 | -2024.0 | -1896.0 | 2.9219 | 1.1172 |
| 0.0169 | 2.5131 | 2400 | 1.7279 | -15.25 | -16.875 | 0.5977 | 1.5625 | -1976.0 | -1848.0 | 2.5156 | 0.6836 |
| 0.0189 | 2.6178 | 2500 | 1.7443 | -15.375 | -17.0 | 0.6016 | 1.6406 | -1992.0 | -1856.0 | 2.6719 | 0.8320 |
| 0.0134 | 2.7225 | 2600 | 1.6854 | -15.0 | -16.625 | 0.5898 | 1.6172 | -1952.0 | -1816.0 | 2.5469 | 0.7109 |
| 0.0161 | 2.8272 | 2700 | 1.6882 | -15.0625 | -16.625 | 0.5879 | 1.625 | -1960.0 | -1824.0 | 2.5312 | 0.6836 |
| 0.0165 | 2.9319 | 2800 | 1.6897 | -15.0625 | -16.625 | 0.5918 | 1.6172 | -1952.0 | -1824.0 | 2.4844 | 0.6445 |
Framework versions
- Transformers 4.44.2
- Pytorch 2.3.0
- Datasets 3.0.0
- Tokenizers 0.19.1