dolly-v2-7b-dpo-full-3-epoch-hydrox-safe
This model is a fine-tuned version of databricks/dolly-v2-7b on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.0060
- Rewards/chosen: 3.6820
- Rewards/rejected: -10.6709
- Rewards/accuracies: 0.9966
- Rewards/margins: 14.3529
- Logps/rejected: -666.2253
- Logps/chosen: -383.1022
- Logits/rejected: -1.3595
- Logits/chosen: -1.5884
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: 4
- seed: 42
- distributed_type: multi-GPU
- num_devices: 8
- total_train_batch_size: 64
- 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: 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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.873 | 0.03 | 100 | 0.7547 | 0.2247 | -0.0238 | 0.5918 | 0.2485 | -559.7548 | -417.6762 | -1.1954 | -1.4929 |
| 0.6069 | 0.07 | 200 | 0.5675 | 0.7273 | -0.0433 | 0.7407 | 0.7706 | -559.9498 | -412.6499 | -1.1932 | -1.4956 |
| 0.3668 | 0.1 | 300 | 0.3913 | 1.3855 | -0.2301 | 0.8552 | 1.6156 | -561.8173 | -406.0676 | -1.1768 | -1.4862 |
| 0.2547 | 0.14 | 400 | 0.2942 | 2.0422 | -0.3359 | 0.8897 | 2.3781 | -562.875 | -399.5007 | -1.1603 | -1.4768 |
| 0.2496 | 0.17 | 500 | 0.2323 | 2.5759 | -0.5597 | 0.9184 | 3.1357 | -565.1138 | -394.1635 | -1.1394 | -1.4661 |
| 0.2099 | 0.2 | 600 | 0.1979 | 3.0353 | -0.7414 | 0.9242 | 3.7767 | -566.9301 | -389.5694 | -1.1137 | -1.4513 |
| 0.123 | 0.24 | 700 | 0.1624 | 3.4398 | -1.1264 | 0.9436 | 4.5662 | -570.7800 | -385.5248 | -1.1147 | -1.4531 |
| 0.1211 | 0.27 | 800 | 0.1404 | 3.7877 | -1.3826 | 0.9453 | 5.1703 | -573.3425 | -382.0456 | -1.1126 | -1.4555 |
| 0.1398 | 0.31 | 900 | 0.1305 | 4.1188 | -1.5720 | 0.9545 | 5.6908 | -575.2359 | -378.7344 | -1.1145 | -1.4568 |
| 0.1161 | 0.34 | 1000 | 0.1066 | 4.3055 | -1.7418 | 0.9646 | 6.0473 | -576.9345 | -376.8678 | -1.1217 | -1.4605 |
| 0.1109 | 0.37 | 1100 | 0.1006 | 4.4233 | -2.0049 | 0.9621 | 6.4282 | -579.5653 | -375.6897 | -1.1334 | -1.4683 |
| 0.0983 | 0.41 | 1200 | 0.0881 | 4.3080 | -2.5628 | 0.9638 | 6.8708 | -585.1442 | -376.8426 | -1.1544 | -1.4813 |
| 0.0965 | 0.44 | 1300 | 0.0778 | 4.3457 | -2.6685 | 0.9621 | 7.0142 | -586.2010 | -376.4656 | -1.1651 | -1.4955 |
| 0.0542 | 0.48 | 1400 | 0.0705 | 4.3768 | -3.1529 | 0.9739 | 7.5297 | -591.0455 | -376.1544 | -1.1767 | -1.4972 |
| 0.053 | 0.51 | 1500 | 0.0659 | 4.4009 | -3.2268 | 0.9781 | 7.6278 | -591.7845 | -375.9133 | -1.1797 | -1.5057 |
| 0.0653 | 0.54 | 1600 | 0.0680 | 4.3566 | -3.1994 | 0.9781 | 7.5559 | -591.5099 | -376.3570 | -1.1682 | -1.4980 |
| 0.0634 | 0.58 | 1700 | 0.0553 | 4.2444 | -3.7967 | 0.9764 | 8.0411 | -597.4832 | -377.4786 | -1.1988 | -1.5176 |
| 0.0574 | 0.61 | 1800 | 0.0490 | 4.3076 | -3.9340 | 0.9790 | 8.2416 | -598.8566 | -376.8465 | -1.2222 | -1.5299 |
| 0.0518 | 0.65 | 1900 | 0.0424 | 4.2820 | -4.0325 | 0.9857 | 8.3145 | -599.8412 | -377.1030 | -1.2285 | -1.5390 |
| 0.0376 | 0.68 | 2000 | 0.0423 | 4.2925 | -4.1097 | 0.9840 | 8.4022 | -600.6129 | -376.9975 | -1.2280 | -1.5311 |
| 0.0339 | 0.71 | 2100 | 0.0424 | 4.2423 | -4.3969 | 0.9882 | 8.6393 | -603.4858 | -377.4996 | -1.2371 | -1.5422 |
| 0.0323 | 0.75 | 2200 | 0.0418 | 4.3016 | -4.3550 | 0.9832 | 8.6566 | -603.0663 | -376.9068 | -1.2198 | -1.5286 |
| 0.0267 | 0.78 | 2300 | 0.0386 | 4.1635 | -4.6663 | 0.9882 | 8.8297 | -606.1791 | -378.2882 | -1.2158 | -1.5230 |
| 0.0296 | 0.82 | 2400 | 0.0316 | 3.9990 | -5.4019 | 0.9907 | 9.4009 | -613.5353 | -379.9330 | -1.2347 | -1.5268 |
| 0.0289 | 0.85 | 2500 | 0.0315 | 4.1064 | -5.2099 | 0.9907 | 9.3164 | -611.6158 | -378.8586 | -1.2152 | -1.5109 |
| 0.0326 | 0.88 | 2600 | 0.0280 | 4.0899 | -5.4030 | 0.9907 | 9.4929 | -613.5463 | -379.0233 | -1.2434 | -1.5354 |
| 0.025 | 0.92 | 2700 | 0.0333 | 4.0463 | -5.2395 | 0.9924 | 9.2857 | -611.9110 | -379.4600 | -1.2283 | -1.5268 |
| 0.0273 | 0.95 | 2800 | 0.0259 | 4.1046 | -5.4947 | 0.9975 | 9.5993 | -614.4639 | -378.8770 | -1.2253 | -1.5271 |
| 0.0197 | 0.99 | 2900 | 0.0360 | 4.1642 | -5.3436 | 0.9907 | 9.5078 | -612.9525 | -378.2808 | -1.2320 | -1.5321 |
| 0.0196 | 1.02 | 3000 | 0.0267 | 3.8748 | -5.9868 | 0.9949 | 9.8616 | -619.3846 | -381.1749 | -1.2358 | -1.5308 |
| 0.0188 | 1.05 | 3100 | 0.0268 | 3.8452 | -6.0908 | 0.9949 | 9.9361 | -620.4247 | -381.4705 | -1.2365 | -1.5361 |
| 0.0172 | 1.09 | 3200 | 0.0231 | 3.7735 | -6.3630 | 0.9907 | 10.1365 | -623.1463 | -382.1877 | -1.2627 | -1.5561 |
| 0.0099 | 1.12 | 3300 | 0.0218 | 3.7491 | -6.5816 | 0.9958 | 10.3307 | -625.3326 | -382.4322 | -1.2410 | -1.5316 |
| 0.0113 | 1.16 | 3400 | 0.0189 | 3.7109 | -6.6907 | 0.9958 | 10.4017 | -626.4235 | -382.8133 | -1.2519 | -1.5387 |
| 0.0146 | 1.19 | 3500 | 0.0191 | 3.6138 | -7.1128 | 0.9941 | 10.7266 | -630.6445 | -383.7852 | -1.2702 | -1.5462 |
| 0.0108 | 1.22 | 3600 | 0.0175 | 3.5940 | -7.3181 | 0.9949 | 10.9121 | -632.6978 | -383.9829 | -1.2642 | -1.5481 |
| 0.0175 | 1.26 | 3700 | 0.0183 | 3.4786 | -7.8254 | 0.9949 | 11.3039 | -637.7700 | -385.1370 | -1.2904 | -1.5503 |
| 0.0147 | 1.29 | 3800 | 0.0153 | 3.2734 | -8.1715 | 0.9966 | 11.4449 | -641.2316 | -387.1888 | -1.3082 | -1.5667 |
| 0.0113 | 1.33 | 3900 | 0.0153 | 3.3033 | -8.3504 | 0.9966 | 11.6537 | -643.0201 | -386.8899 | -1.2907 | -1.5525 |
| 0.0284 | 1.36 | 4000 | 0.0270 | 3.5241 | -8.1571 | 0.9924 | 11.6812 | -641.0871 | -384.6817 | -1.2917 | -1.5474 |
| 0.0101 | 1.39 | 4100 | 0.0138 | 3.3142 | -8.9443 | 0.9941 | 12.2585 | -648.9590 | -386.7809 | -1.3039 | -1.5402 |
| 0.0093 | 1.43 | 4200 | 0.0159 | 3.3533 | -9.0499 | 0.9966 | 12.4032 | -650.0153 | -386.3899 | -1.3067 | -1.5543 |
| 0.0083 | 1.46 | 4300 | 0.0149 | 3.4209 | -8.8296 | 0.9958 | 12.2505 | -647.8128 | -385.7142 | -1.3104 | -1.5558 |
| 0.0068 | 1.5 | 4400 | 0.0123 | 3.2700 | -9.3033 | 0.9975 | 12.5733 | -652.5496 | -387.2229 | -1.3257 | -1.5680 |
| 0.0093 | 1.53 | 4500 | 0.0122 | 3.5894 | -8.8354 | 0.9983 | 12.4248 | -647.8701 | -384.0288 | -1.3217 | -1.5701 |
| 0.0065 | 1.56 | 4600 | 0.0117 | 3.4515 | -8.7814 | 0.9975 | 12.2329 | -647.3306 | -385.4080 | -1.3381 | -1.5838 |
| 0.0132 | 1.6 | 4700 | 0.0119 | 3.4540 | -8.4518 | 0.9975 | 11.9058 | -644.0345 | -385.3825 | -1.3352 | -1.5862 |
| 0.0085 | 1.63 | 4800 | 0.0113 | 3.3970 | -8.7353 | 0.9966 | 12.1323 | -646.8692 | -385.9526 | -1.3331 | -1.5766 |
| 0.0096 | 1.67 | 4900 | 0.0121 | 3.2728 | -9.0713 | 0.9966 | 12.3442 | -650.2295 | -387.1943 | -1.3552 | -1.5969 |
| 0.0042 | 1.7 | 5000 | 0.0106 | 3.1699 | -9.4193 | 0.9975 | 12.5892 | -653.7093 | -388.2237 | -1.3307 | -1.5739 |
| 0.0116 | 1.73 | 5100 | 0.0096 | 3.2716 | -9.0292 | 0.9958 | 12.3008 | -649.8085 | -387.2067 | -1.3274 | -1.5748 |
| 0.0093 | 1.77 | 5200 | 0.0103 | 3.2228 | -9.3477 | 0.9983 | 12.5706 | -652.9938 | -387.6946 | -1.3153 | -1.5495 |
| 0.0058 | 1.8 | 5300 | 0.0103 | 3.1251 | -9.6052 | 0.9966 | 12.7303 | -655.5681 | -388.6714 | -1.3273 | -1.5594 |
| 0.0066 | 1.84 | 5400 | 0.0094 | 3.5167 | -9.0559 | 0.9983 | 12.5726 | -650.0754 | -384.7553 | -1.3330 | -1.5721 |
| 0.0038 | 1.87 | 5500 | 0.0093 | 3.5884 | -9.0262 | 0.9983 | 12.6146 | -649.7783 | -384.0386 | -1.3171 | -1.5599 |
| 0.0134 | 1.9 | 5600 | 0.0093 | 3.0874 | -9.8027 | 0.9983 | 12.8901 | -657.5432 | -389.0488 | -1.3368 | -1.5645 |
| 0.0059 | 1.94 | 5700 | 0.0098 | 3.4393 | -9.7104 | 0.9975 | 13.1497 | -656.6204 | -385.5294 | -1.3526 | -1.5716 |
| 0.0057 | 1.97 | 5800 | 0.0080 | 3.5892 | -9.4003 | 0.9983 | 12.9896 | -653.5198 | -384.0307 | -1.3593 | -1.5880 |
| 0.0015 | 2.01 | 5900 | 0.0102 | 3.4266 | -9.8551 | 0.9966 | 13.2816 | -658.0669 | -385.6569 | -1.3552 | -1.5837 |
| 0.0019 | 2.04 | 6000 | 0.0105 | 3.5092 | -9.9457 | 0.9983 | 13.4549 | -658.9734 | -384.8311 | -1.3418 | -1.5734 |
| 0.0049 | 2.07 | 6100 | 0.0083 | 3.4872 | -10.1039 | 0.9983 | 13.5911 | -660.5549 | -385.0504 | -1.3269 | -1.5633 |
| 0.0056 | 2.11 | 6200 | 0.0089 | 3.3922 | -10.3713 | 0.9975 | 13.7635 | -663.2297 | -386.0008 | -1.3437 | -1.5700 |
| 0.0041 | 2.14 | 6300 | 0.0078 | 3.5705 | -10.1344 | 0.9983 | 13.7049 | -660.8607 | -384.2182 | -1.3527 | -1.5831 |
| 0.0039 | 2.18 | 6400 | 0.0092 | 3.3798 | -10.7994 | 0.9975 | 14.1792 | -667.5103 | -386.1252 | -1.3748 | -1.5843 |
| 0.0018 | 2.21 | 6500 | 0.0076 | 3.5825 | -10.5328 | 0.9983 | 14.1153 | -664.8441 | -384.0977 | -1.3583 | -1.5744 |
| 0.0037 | 2.24 | 6600 | 0.0075 | 3.5553 | -10.3432 | 0.9983 | 13.8984 | -662.9481 | -384.3702 | -1.3604 | -1.5848 |
| 0.0021 | 2.28 | 6700 | 0.0082 | 3.7310 | -10.3324 | 0.9992 | 14.0634 | -662.8404 | -382.6127 | -1.3437 | -1.5693 |
| 0.0025 | 2.31 | 6800 | 0.0074 | 3.5582 | -10.6710 | 0.9975 | 14.2292 | -666.2263 | -384.3409 | -1.3487 | -1.5658 |
| 0.0112 | 2.35 | 6900 | 0.0076 | 3.5915 | -10.7786 | 0.9966 | 14.3700 | -667.3019 | -384.0081 | -1.3470 | -1.5688 |
| 0.0022 | 2.38 | 7000 | 0.0080 | 3.6060 | -10.6007 | 0.9975 | 14.2067 | -665.5234 | -383.8625 | -1.3536 | -1.5774 |
| 0.0012 | 2.41 | 7100 | 0.0063 | 3.5627 | -10.8773 | 0.9975 | 14.4400 | -668.2891 | -384.2953 | -1.3445 | -1.5681 |
| 0.0018 | 2.45 | 7200 | 0.0070 | 3.4237 | -11.0692 | 0.9975 | 14.4928 | -670.2083 | -385.6862 | -1.3656 | -1.5819 |
| 0.0084 | 2.48 | 7300 | 0.0079 | 3.7091 | -10.3477 | 0.9983 | 14.0569 | -662.9936 | -382.8314 | -1.3539 | -1.5873 |
| 0.0031 | 2.52 | 7400 | 0.0064 | 3.5680 | -10.4848 | 0.9983 | 14.0528 | -664.3639 | -384.2423 | -1.3510 | -1.5829 |
| 0.0027 | 2.55 | 7500 | 0.0069 | 3.5130 | -10.6612 | 0.9983 | 14.1741 | -666.1280 | -384.7932 | -1.3666 | -1.5947 |
| 0.0051 | 2.58 | 7600 | 0.0066 | 3.5461 | -10.7595 | 0.9983 | 14.3056 | -667.1109 | -384.4612 | -1.3600 | -1.5872 |
| 0.001 | 2.62 | 7700 | 0.0076 | 3.5633 | -10.7690 | 0.9983 | 14.3323 | -667.2067 | -384.2903 | -1.3486 | -1.5750 |
| 0.0021 | 2.65 | 7800 | 0.0066 | 3.6662 | -10.7670 | 0.9983 | 14.4332 | -667.1862 | -383.2607 | -1.3604 | -1.5892 |
| 0.004 | 2.69 | 7900 | 0.0067 | 3.7915 | -10.4856 | 0.9983 | 14.2771 | -664.3723 | -382.0074 | -1.3540 | -1.5830 |
| 0.0022 | 2.72 | 8000 | 0.0066 | 3.8259 | -10.5371 | 0.9983 | 14.3630 | -664.8873 | -381.6641 | -1.3510 | -1.5812 |
| 0.0018 | 2.75 | 8100 | 0.0071 | 3.7228 | -10.6783 | 0.9983 | 14.4011 | -666.2990 | -382.6946 | -1.3470 | -1.5789 |
| 0.0015 | 2.79 | 8200 | 0.0065 | 3.7032 | -10.7685 | 0.9983 | 14.4717 | -667.2010 | -382.8909 | -1.3501 | -1.5791 |
| 0.0015 | 2.82 | 8300 | 0.0072 | 3.7173 | -10.7747 | 0.9975 | 14.4920 | -667.2634 | -382.7499 | -1.3574 | -1.5864 |
| 0.0016 | 2.86 | 8400 | 0.0064 | 3.7268 | -10.7100 | 0.9983 | 14.4368 | -666.6169 | -382.6550 | -1.3626 | -1.5899 |
| 0.0022 | 2.89 | 8500 | 0.0062 | 3.6175 | -10.9238 | 0.9992 | 14.5413 | -668.7542 | -383.7477 | -1.3531 | -1.5792 |
| 0.0032 | 2.92 | 8600 | 0.0059 | 3.7174 | -10.7382 | 0.9983 | 14.4556 | -666.8983 | -382.7484 | -1.3576 | -1.5869 |
| 0.0035 | 2.96 | 8700 | 0.0062 | 3.5749 | -11.0859 | 0.9975 | 14.6608 | -670.3754 | -384.1739 | -1.3459 | -1.5728 |
| 0.0024 | 2.99 | 8800 | 0.0062 | 3.6894 | -10.8042 | 0.9983 | 14.4936 | -667.5587 | -383.0290 | -1.3618 | -1.5891 |
Framework versions
- Transformers 4.35.0
- Pytorch 2.1.1+cu121
- Datasets 2.14.6
- Tokenizers 0.14.1
- Downloads last month
- 4
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.
Model tree for yihang7/dolly-v2-7b-dpo-full-3-epoch-hydrox-safe
Base model
databricks/dolly-v2-7b