metadata
library_name: transformers
license: llama3
base_model: tsavage68/IE_L3_1000steps_1e6rate_SFT
tags:
- trl
- dpo
- generated_from_trainer
model-index:
- name: IE_L3_1000steps_1e7rate_01beta_cSFTDPO
results: []
IE_L3_1000steps_1e7rate_01beta_cSFTDPO
This model is a fine-tuned version of tsavage68/IE_L3_1000steps_1e6rate_SFT on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.1803
- Rewards/chosen: -0.5346
- Rewards/rejected: -8.6468
- Rewards/accuracies: 0.7400
- Rewards/margins: 8.1123
- Logps/rejected: -162.0956
- Logps/chosen: -88.1433
- Logits/rejected: -0.8498
- Logits/chosen: -0.7319
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: 1e-07
- train_batch_size: 2
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 4
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 100
- training_steps: 1000
Training results
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.6292 | 0.4 | 50 | 0.5972 | -0.0178 | -0.2247 | 0.7400 | 0.2070 | -77.8745 | -82.9754 | -0.7952 | -0.7369 |
| 0.2432 | 0.8 | 100 | 0.2531 | -0.1984 | -1.9084 | 0.7400 | 1.7099 | -94.7109 | -84.7823 | -0.7935 | -0.7222 |
| 0.1468 | 1.2 | 150 | 0.1842 | -0.4156 | -4.4900 | 0.7400 | 4.0744 | -120.5273 | -86.9542 | -0.8149 | -0.7193 |
| 0.1745 | 1.6 | 200 | 0.1807 | -0.4305 | -6.5857 | 0.7400 | 6.1551 | -141.4839 | -87.1031 | -0.8342 | -0.7283 |
| 0.2254 | 2.0 | 250 | 0.1805 | -0.4554 | -7.3110 | 0.7400 | 6.8555 | -148.7368 | -87.3519 | -0.8373 | -0.7278 |
| 0.1389 | 2.4 | 300 | 0.1804 | -0.4666 | -7.7073 | 0.7400 | 7.2408 | -152.7006 | -87.4635 | -0.8397 | -0.7280 |
| 0.1215 | 2.8 | 350 | 0.1804 | -0.4933 | -8.0779 | 0.7400 | 7.5846 | -156.4058 | -87.7304 | -0.8446 | -0.7309 |
| 0.191 | 3.2 | 400 | 0.1804 | -0.5121 | -8.2398 | 0.7400 | 7.7277 | -158.0253 | -87.9188 | -0.8463 | -0.7322 |
| 0.1906 | 3.6 | 450 | 0.1804 | -0.5199 | -8.2886 | 0.7400 | 7.7687 | -158.5128 | -87.9963 | -0.8471 | -0.7317 |
| 0.2084 | 4.0 | 500 | 0.1804 | -0.5104 | -8.4325 | 0.7400 | 7.9221 | -159.9520 | -87.9018 | -0.8488 | -0.7326 |
| 0.1561 | 4.4 | 550 | 0.1803 | -0.5293 | -8.5197 | 0.7400 | 7.9905 | -160.8244 | -88.0903 | -0.8493 | -0.7326 |
| 0.1213 | 4.8 | 600 | 0.1803 | -0.5356 | -8.5680 | 0.7400 | 8.0324 | -161.3075 | -88.1538 | -0.8503 | -0.7332 |
| 0.1907 | 5.2 | 650 | 0.1803 | -0.5333 | -8.6184 | 0.7400 | 8.0851 | -161.8111 | -88.1307 | -0.8505 | -0.7330 |
| 0.2427 | 5.6 | 700 | 0.1803 | -0.5362 | -8.6233 | 0.7400 | 8.0871 | -161.8604 | -88.1602 | -0.8507 | -0.7332 |
| 0.2601 | 6.0 | 750 | 0.1803 | -0.5367 | -8.6352 | 0.7400 | 8.0985 | -161.9794 | -88.1651 | -0.8509 | -0.7332 |
| 0.1213 | 6.4 | 800 | 0.1803 | -0.5353 | -8.6312 | 0.7400 | 8.0960 | -161.9397 | -88.1506 | -0.8507 | -0.7334 |
| 0.2426 | 6.8 | 850 | 0.1803 | -0.5305 | -8.6468 | 0.7400 | 8.1163 | -162.0951 | -88.1023 | -0.8507 | -0.7328 |
| 0.1733 | 7.2 | 900 | 0.1803 | -0.5246 | -8.6359 | 0.7400 | 8.1112 | -161.9858 | -88.0442 | -0.8503 | -0.7323 |
| 0.1388 | 7.6 | 950 | 0.1803 | -0.5346 | -8.6468 | 0.7400 | 8.1123 | -162.0956 | -88.1433 | -0.8498 | -0.7319 |
| 0.1561 | 8.0 | 1000 | 0.1803 | -0.5346 | -8.6468 | 0.7400 | 8.1123 | -162.0956 | -88.1433 | -0.8498 | -0.7319 |
Framework versions
- Transformers 4.44.2
- Pytorch 2.0.0+cu117
- Datasets 3.0.0
- Tokenizers 0.19.1