File size: 13,224 Bytes
149e77a 2089607 149e77a 2089607 149e77a 2089607 149e77a ea4ba97 2089607 149e77a ea4ba97 149e77a |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 |
---
license: apache-2.0
library_name: peft
tags:
- alignment-handbook
- trl
- dpo
- generated_from_trainer
base_model: TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T
datasets:
- openai/summarize_from_feedback
model-index:
- name: tinyllama-1.1b-sum-dpo-qlora
results: []
---
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# tinyllama-1.1b-sum-dpo-qlora
This model is a fine-tuned version of [martimfasantos/tinyllama-1.1b-sum-sft-qlora](https://huggingface.co/martimfasantos/tinyllama-1.1b-sum-sft-qlora) on the openai/summarize_from_feedback dataset.
It achieves the following results on the evaluation set:
- Loss: 0.6482
- Rewards/chosen: -0.9538
- Rewards/rejected: -1.1194
- Rewards/accuracies: 0.6171
- Rewards/margins: 0.1656
- Logps/rejected: -187.0472
- Logps/chosen: -166.7881
- Logits/rejected: -3.0176
- Logits/chosen: -3.0239
## 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-06
- train_batch_size: 4
- eval_batch_size: 8
- seed: 42
- distributed_type: multi-GPU
- gradient_accumulation_steps: 4
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 1
### Training results
| Training Loss | Epoch | Step | Logits/chosen | Logits/rejected | Logps/chosen | Logps/rejected | Validation Loss | Rewards/accuracies | Rewards/chosen | Rewards/margins | Rewards/rejected |
|:-------------:|:-----:|:----:|:-------------:|:---------------:|:------------:|:--------------:|:---------------:|:------------------:|:--------------:|:---------------:|:----------------:|
| 0.6926 | 0.02 | 100 | -3.4980 | -3.4962 | -70.9186 | -74.6392 | 0.6930 | 0.5193 | 0.0049 | 0.0002 | 0.0047 |
| 0.6919 | 0.03 | 200 | -3.4925 | -3.4908 | -69.9505 | -73.7540 | 0.6926 | 0.5678 | 0.0146 | 0.0011 | 0.0135 |
| 0.6888 | 0.05 | 300 | -3.4861 | -3.4843 | -67.8994 | -72.0238 | 0.6911 | 0.5748 | 0.0351 | 0.0043 | 0.0308 |
| 0.6864 | 0.07 | 400 | -3.4827 | -3.4809 | -69.7504 | -74.3218 | 0.6890 | 0.5627 | 0.0166 | 0.0087 | 0.0079 |
| 0.6864 | 0.09 | 500 | -3.4687 | -3.4669 | -69.0559 | -74.2092 | 0.6864 | 0.5716 | 0.0235 | 0.0146 | 0.0090 |
| 0.6729 | 0.1 | 600 | -3.4506 | -3.4489 | -71.3562 | -77.1629 | 0.6837 | 0.5869 | 0.0005 | 0.0211 | -0.0206 |
| 0.6745 | 0.12 | 700 | -3.4487 | -3.4467 | -78.9372 | -85.9956 | 0.6786 | 0.5955 | -0.0753 | 0.0336 | -0.1089 |
| 0.6681 | 0.14 | 800 | -3.4169 | -3.4151 | -90.1915 | -98.6570 | 0.6738 | 0.5955 | -0.1878 | 0.0477 | -0.2355 |
| 0.6661 | 0.16 | 900 | -3.3755 | -3.3740 | -88.5994 | -97.6376 | 0.6715 | 0.5922 | -0.1719 | 0.0534 | -0.2253 |
| 0.6686 | 0.17 | 1000 | -3.3483 | -3.3467 | -111.1606 | -121.9167 | 0.6681 | 0.5936 | -0.3975 | 0.0706 | -0.4681 |
| 0.665 | 0.19 | 1100 | -3.3477 | -3.3463 | -92.1750 | -101.6747 | 0.6708 | 0.5950 | -0.2076 | 0.0580 | -0.2657 |
| 0.6549 | 0.21 | 1200 | -3.3173 | -3.3159 | -107.3321 | -119.3906 | 0.6631 | 0.5974 | -0.3592 | 0.0836 | -0.4428 |
| 0.6536 | 0.22 | 1300 | -3.2737 | -3.2722 | -121.8111 | -135.5439 | 0.6591 | 0.5978 | -0.5040 | 0.1004 | -0.6044 |
| 0.6303 | 0.24 | 1400 | -3.2790 | -3.2775 | -111.6529 | -124.7296 | 0.6593 | 0.6055 | -0.4024 | 0.0938 | -0.4962 |
| 0.6611 | 0.26 | 1500 | -3.2472 | -3.2454 | -132.2458 | -148.1280 | 0.6527 | 0.6138 | -0.6084 | 0.1219 | -0.7302 |
| 0.6395 | 0.28 | 1600 | -3.2525 | -3.2505 | -126.2706 | -141.6170 | 0.6536 | 0.6155 | -0.5486 | 0.1165 | -0.6651 |
| 0.678 | 0.29 | 1700 | -3.2125 | -3.2107 | -117.8728 | -131.2285 | 0.6587 | 0.6169 | -0.4646 | 0.0966 | -0.5612 |
| 0.629 | 0.31 | 1800 | -3.1113 | -3.1087 | -146.8860 | -164.9026 | 0.6489 | 0.6187 | -0.7548 | 0.1432 | -0.8980 |
| 0.6622 | 0.33 | 1900 | -3.1419 | -3.1399 | -125.9992 | -140.6700 | 0.6555 | 0.6069 | -0.5459 | 0.1097 | -0.6556 |
| 0.64 | 0.34 | 2000 | -3.1847 | -3.1824 | -140.1714 | -156.3843 | 0.6523 | 0.6101 | -0.6876 | 0.1252 | -0.8128 |
| 0.6479 | 0.36 | 2100 | -3.1160 | -3.1130 | -150.8988 | -167.6336 | 0.6537 | 0.6104 | -0.7949 | 0.1304 | -0.9253 |
| 0.6023 | 0.38 | 2200 | -3.1479 | -3.1449 | -137.7163 | -153.7927 | 0.6536 | 0.6034 | -0.6631 | 0.1238 | -0.7869 |
| 0.5962 | 0.4 | 2300 | -3.1012 | -3.0975 | -159.4141 | -177.2301 | 0.6523 | 0.6078 | -0.8800 | 0.1412 | -1.0212 |
| 0.6176 | 0.41 | 2400 | -3.0320 | -3.0265 | -172.7089 | -192.7748 | 0.6506 | 0.6027 | -1.0130 | 0.1637 | -1.1767 |
| 0.6255 | 0.43 | 2500 | -3.0629 | -3.0584 | -156.9642 | -175.3398 | 0.6507 | 0.6101 | -0.8555 | 0.1468 | -1.0023 |
| 0.6075 | 0.45 | 2600 | -3.0877 | -3.0839 | -146.0736 | -162.3147 | 0.6547 | 0.6046 | -0.7466 | 0.1254 | -0.8721 |
| 0.6282 | 0.47 | 2700 | -3.1221 | -3.1185 | -140.7325 | -157.2624 | 0.6531 | 0.6101 | -0.6932 | 0.1283 | -0.8216 |
| 0.6495 | 0.48 | 2800 | -3.0926 | -3.0887 | -148.7372 | -166.3009 | 0.6517 | 0.6080 | -0.7733 | 0.1387 | -0.9119 |
| 0.6202 | 0.5 | 2900 | -3.0787 | -3.0744 | -152.9659 | -170.9832 | 0.6512 | 0.6048 | -0.8156 | 0.1432 | -0.9588 |
| 0.6252 | 0.52 | 3000 | -3.0824 | -3.0782 | -148.4267 | -166.3868 | 0.6505 | 0.6055 | -0.7702 | 0.1426 | -0.9128 |
| 0.6082 | 0.53 | 3100 | -3.0723 | -3.0678 | -149.2047 | -167.4548 | 0.6500 | 0.6115 | -0.7779 | 0.1455 | -0.9235 |
| 0.6072 | 0.55 | 3200 | -3.0863 | -3.0819 | -147.0810 | -164.9669 | 0.6499 | 0.6090 | -0.7567 | 0.1419 | -0.8986 |
| 0.6142 | 0.57 | 3300 | -3.0087 | -3.0026 | -179.2665 | -200.5992 | 0.6468 | 0.6176 | -1.0786 | 0.1764 | -1.2549 |
| 0.602 | 0.59 | 3400 | -3.0674 | -3.0624 | -150.3082 | -168.4087 | 0.6504 | 0.6136 | -0.7890 | 0.1440 | -0.9330 |
| 0.605 | 0.6 | 3500 | -3.0590 | -3.0538 | -154.1790 | -172.9109 | 0.6497 | 0.6122 | -0.8277 | 0.1503 | -0.9780 |
| 0.6263 | 0.62 | 3600 | -3.0721 | -3.0672 | -149.9757 | -168.0735 | 0.6508 | 0.6043 | -0.7857 | 0.1440 | -0.9297 |
| 0.5961 | 0.64 | 3700 | -3.0151 | -3.0090 | -169.4567 | -189.3689 | 0.6492 | 0.6136 | -0.9805 | 0.1622 | -1.1426 |
| 0.6273 | 0.65 | 3800 | -3.0117 | -3.0057 | -167.9805 | -187.6573 | 0.6494 | 0.6141 | -0.9657 | 0.1598 | -1.1255 |
| 0.6183 | 0.67 | 3900 | -3.0137 | -3.0077 | -167.4417 | -187.2734 | 0.6488 | 0.6166 | -0.9603 | 0.1613 | -1.1217 |
| 0.6051 | 0.69 | 4000 | -2.9974 | -2.9908 | -176.3739 | -197.1255 | 0.6482 | 0.6178 | -1.0496 | 0.1705 | -1.2202 |
| 0.5867 | 0.71 | 4100 | -3.0151 | -3.0088 | -169.1084 | -189.3998 | 0.6484 | 0.6125 | -0.9770 | 0.1659 | -1.1429 |
| 0.6554 | 0.72 | 4200 | -3.0270 | -3.0209 | -164.2755 | -184.0126 | 0.6489 | 0.6176 | -0.9287 | 0.1604 | -1.0891 |
| 0.6053 | 0.74 | 4300 | -3.0362 | -3.0303 | -159.9774 | -179.4446 | 0.6489 | 0.6097 | -0.8857 | 0.1577 | -1.0434 |
| 0.6153 | 0.76 | 4400 | -3.0351 | -3.0292 | -160.5470 | -180.1235 | 0.6489 | 0.6120 | -0.8914 | 0.1588 | -1.0502 |
| 0.6145 | 0.78 | 4500 | -3.0378 | -3.0319 | -160.1720 | -179.6728 | 0.6490 | 0.6113 | -0.8876 | 0.1580 | -1.0457 |
| 0.5798 | 0.79 | 4600 | -3.0308 | -3.0247 | -162.6813 | -182.4701 | 0.6488 | 0.6148 | -0.9127 | 0.1609 | -1.0736 |
| 0.6218 | 0.81 | 4700 | -3.0307 | -3.0246 | -163.0493 | -182.9482 | 0.6486 | 0.6152 | -0.9164 | 0.1620 | -1.0784 |
| 0.6102 | 0.83 | 4800 | -3.0259 | -3.0197 | -164.8939 | -184.9769 | 0.6484 | 0.6150 | -0.9348 | 0.1639 | -1.0987 |
| 0.6176 | 0.84 | 4900 | -3.0273 | -3.0211 | -165.7554 | -185.9428 | 0.6483 | 0.6157 | -0.9435 | 0.1649 | -1.1084 |
| 0.5907 | 0.86 | 5000 | -3.0259 | -3.0196 | -167.1301 | -187.4627 | 0.6482 | 0.6164 | -0.9572 | 0.1664 | -1.1236 |
| 0.6534 | 0.88 | 5100 | -3.0211 | -3.0148 | -167.2241 | -187.5712 | 0.6481 | 0.6155 | -0.9581 | 0.1665 | -1.1246 |
| 0.5973 | 0.9 | 5200 | -3.0194 | -3.0130 | -166.8823 | -187.1679 | 0.6483 | 0.6169 | -0.9547 | 0.1659 | -1.1206 |
| 0.5975 | 0.91 | 5300 | -3.0248 | -3.0185 | -166.6118 | -186.8759 | 0.6482 | 0.6162 | -0.9520 | 0.1657 | -1.1177 |
| 0.5986 | 0.93 | 5400 | -3.0249 | -3.0186 | -166.6502 | -186.8928 | 0.6483 | 0.6190 | -0.9524 | 0.1655 | -1.1179 |
| 0.6025 | 0.95 | 5500 | -3.0252 | -3.0189 | -166.7467 | -186.9980 | 0.6483 | 0.6169 | -0.9534 | 0.1655 | -1.1189 |
| 0.6149 | 0.96 | 5600 | -3.0244 | -3.0181 | -166.7859 | -187.1137 | 0.6480 | 0.6155 | -0.9538 | 0.1663 | -1.1201 |
| 0.6275 | 0.98 | 5700 | -3.0245 | -3.0182 | -166.6791 | -186.9484 | 0.6482 | 0.6178 | -0.9527 | 0.1657 | -1.1184 |
| 0.5876 | 1.0 | 5800 | -3.0239 | -3.0176 | -166.7881 | -187.0472 | 0.6482 | 0.6171 | -0.9538 | 0.1656 | -1.1194 |
### Framework versions
- PEFT 0.7.1
- Transformers 4.39.3
- Pytorch 2.1.2
- Datasets 2.18.0
- Tokenizers 0.15.2 |