--- library_name: transformers tags: - trl - dpo - alignment-handbook - generated_from_trainer model-index: - name: OpenELM-1_1B-DPO-full-max-reward-most-similar results: [] --- # OpenELM-1_1B-DPO-full-max-reward-most-similar This model was trained from scratch on an unknown dataset. It achieves the following results on the evaluation set: - Loss: 1.6250 - Rewards/chosen: -16.375 - Rewards/rejected: -18.625 - Rewards/accuracies: 0.6055 - Rewards/margins: 2.1406 - Logps/rejected: -2144.0 - Logps/chosen: -1960.0 - Logits/rejected: 4.75 - Logits/chosen: 3.2969 ## 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 | |:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:| | 0.5787 | 0.1047 | 100 | 0.6727 | -1.3594 | -1.6328 | 0.6133 | 0.2793 | -452.0 | -454.0 | -10.4375 | -10.875 | | 0.5561 | 0.2094 | 200 | 0.7091 | -2.5625 | -3.0625 | 0.6387 | 0.5039 | -596.0 | -576.0 | -3.2969 | -4.3438 | | 0.5165 | 0.3141 | 300 | 0.7539 | -4.5938 | -5.1875 | 0.6035 | 0.6016 | -808.0 | -776.0 | -2.8438 | -3.9844 | | 0.5008 | 0.4188 | 400 | 0.7994 | -4.7188 | -5.3438 | 0.5918 | 0.6094 | -824.0 | -792.0 | -3.5469 | -4.8438 | | 0.505 | 0.5236 | 500 | 0.8047 | -6.2188 | -6.9375 | 0.6016 | 0.7109 | -980.0 | -940.0 | -1.7812 | -3.1406 | | 0.5125 | 0.6283 | 600 | 0.8067 | -5.2812 | -5.9375 | 0.6133 | 0.6562 | -884.0 | -848.0 | -1.3281 | -2.6562 | | 0.4627 | 0.7330 | 700 | 0.9020 | -4.9375 | -5.9062 | 0.6172 | 0.9648 | -880.0 | -812.0 | -2.1719 | -3.7969 | | 0.4583 | 0.8377 | 800 | 0.8516 | -7.3125 | -7.9688 | 0.5488 | 0.6562 | -1088.0 | -1048.0 | -0.8438 | -2.25 | | 0.4454 | 0.9424 | 900 | 0.9168 | -6.6875 | -7.5938 | 0.6094 | 0.9297 | -1048.0 | -984.0 | -3.4375 | -5.0 | | 0.1656 | 1.0471 | 1000 | 1.0883 | -9.1875 | -10.3125 | 0.6055 | 1.1172 | -1320.0 | -1240.0 | 0.2676 | -1.3984 | | 0.1613 | 1.1518 | 1100 | 1.0450 | -9.125 | -10.25 | 0.6152 | 1.1797 | -1312.0 | -1232.0 | 1.8438 | 0.1934 | | 0.1616 | 1.2565 | 1200 | 1.0938 | -10.0625 | -11.1875 | 0.5996 | 1.1484 | -1408.0 | -1320.0 | 0.8789 | -0.5898 | | 0.1677 | 1.3613 | 1300 | 0.9728 | -7.6562 | -8.9375 | 0.6504 | 1.2422 | -1184.0 | -1088.0 | -0.3652 | -1.9297 | | 0.1712 | 1.4660 | 1400 | 1.0911 | -9.8125 | -11.0625 | 0.6016 | 1.2266 | -1392.0 | -1296.0 | -0.3105 | -1.7344 | | 0.1723 | 1.5707 | 1500 | 1.0245 | -10.0 | -11.25 | 0.6172 | 1.3047 | -1416.0 | -1312.0 | 0.6992 | -0.7773 | | 0.1515 | 1.6754 | 1600 | 1.0037 | -9.0 | -10.3125 | 0.6211 | 1.3359 | -1320.0 | -1216.0 | -0.1299 | -1.7109 | | 0.1197 | 1.7801 | 1700 | 1.1304 | -10.75 | -12.125 | 0.5977 | 1.3594 | -1504.0 | -1392.0 | 0.9805 | -0.4961 | | 0.1258 | 1.8848 | 1800 | 1.0837 | -10.5 | -11.8125 | 0.6113 | 1.3359 | -1472.0 | -1368.0 | 2.2188 | 0.6797 | | 0.118 | 1.9895 | 1900 | 1.0529 | -11.5 | -12.75 | 0.6172 | 1.2578 | -1560.0 | -1464.0 | 3.4062 | 2.0 | | 0.03 | 2.0942 | 2000 | 1.4271 | -14.75 | -16.625 | 0.5977 | 1.8516 | -1952.0 | -1792.0 | 4.4375 | 3.0 | | 0.0206 | 2.1990 | 2100 | 1.4637 | -14.75 | -16.75 | 0.6230 | 2.0469 | -1968.0 | -1792.0 | 4.8125 | 3.3594 | | 0.0301 | 2.3037 | 2200 | 1.5228 | -15.4375 | -17.5 | 0.6152 | 2.0469 | -2040.0 | -1864.0 | 4.625 | 3.1875 | | 0.0312 | 2.4084 | 2300 | 1.6112 | -16.375 | -18.5 | 0.6172 | 2.0781 | -2128.0 | -1952.0 | 4.75 | 3.3125 | | 0.0379 | 2.5131 | 2400 | 1.6100 | -16.5 | -18.625 | 0.6113 | 2.1094 | -2144.0 | -1968.0 | 4.5938 | 3.1719 | | 0.0235 | 2.6178 | 2500 | 1.6536 | -17.0 | -19.125 | 0.6113 | 2.1719 | -2208.0 | -2016.0 | 4.75 | 3.3125 | | 0.0198 | 2.7225 | 2600 | 1.6602 | -16.75 | -19.0 | 0.6113 | 2.2031 | -2192.0 | -1992.0 | 4.8125 | 3.3594 | | 0.0251 | 2.8272 | 2700 | 1.6070 | -16.25 | -18.375 | 0.6133 | 2.1406 | -2128.0 | -1936.0 | 4.7188 | 3.2656 | | 0.0213 | 2.9319 | 2800 | 1.6250 | -16.375 | -18.625 | 0.6055 | 2.1406 | -2144.0 | -1960.0 | 4.75 | 3.2969 | ### Framework versions - Transformers 4.44.2 - Pytorch 2.3.0 - Datasets 3.0.0 - Tokenizers 0.19.1