Na_L3_1000steps_1e6rate_05beta_cSFTDPO
This model is a fine-tuned version of tsavage68/Na_L3_100steps_1e6rate_SFT on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.0000
- Rewards/chosen: 1.8884
- Rewards/rejected: -13.4648
- Rewards/accuracies: 1.0
- Rewards/margins: 15.3532
- Logps/rejected: -68.4313
- Logps/chosen: -21.1134
- Logits/rejected: -0.9549
- Logits/chosen: -0.8862
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-06
- 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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.0 | 0.2667 | 50 | 0.0000 | 1.3180 | -9.6006 | 1.0 | 10.9185 | -60.7028 | -22.2542 | -0.9556 | -0.8883 |
| 0.0 | 0.5333 | 100 | 0.0000 | 1.5811 | -11.1723 | 1.0 | 12.7534 | -63.8463 | -21.7279 | -0.9556 | -0.8877 |
| 0.0 | 0.8 | 150 | 0.0000 | 1.7220 | -11.8689 | 1.0 | 13.5909 | -65.2395 | -21.4461 | -0.9552 | -0.8869 |
| 0.0 | 1.0667 | 200 | 0.0000 | 1.7657 | -12.2278 | 1.0 | 13.9935 | -65.9574 | -21.3588 | -0.9548 | -0.8865 |
| 0.0 | 1.3333 | 250 | 0.0000 | 1.7855 | -12.5076 | 1.0 | 14.2932 | -66.5170 | -21.3191 | -0.9541 | -0.8858 |
| 0.0 | 1.6 | 300 | 0.0000 | 1.8162 | -12.7234 | 1.0 | 14.5396 | -66.9486 | -21.2578 | -0.9548 | -0.8864 |
| 0.0 | 1.8667 | 350 | 0.0000 | 1.8322 | -12.8906 | 1.0 | 14.7228 | -67.2829 | -21.2257 | -0.9552 | -0.8868 |
| 0.0 | 2.1333 | 400 | 0.0000 | 1.8447 | -13.0321 | 1.0 | 14.8768 | -67.5660 | -21.2008 | -0.9541 | -0.8855 |
| 0.0 | 2.4 | 450 | 0.0000 | 1.8576 | -13.1430 | 1.0 | 15.0006 | -67.7877 | -21.1749 | -0.9542 | -0.8856 |
| 0.0 | 2.6667 | 500 | 0.0000 | 1.8705 | -13.2358 | 1.0 | 15.1063 | -67.9734 | -21.1491 | -0.9535 | -0.8847 |
| 0.0 | 2.9333 | 550 | 0.0000 | 1.8928 | -13.3108 | 1.0 | 15.2037 | -68.1234 | -21.1045 | -0.9551 | -0.8865 |
| 0.0 | 3.2 | 600 | 0.0000 | 1.8716 | -13.3744 | 1.0 | 15.2461 | -68.2506 | -21.1468 | -0.9545 | -0.8858 |
| 0.0 | 3.4667 | 650 | 0.0000 | 1.8898 | -13.4154 | 1.0 | 15.3052 | -68.3326 | -21.1106 | -0.9551 | -0.8865 |
| 0.0 | 3.7333 | 700 | 0.0000 | 1.8784 | -13.4316 | 1.0 | 15.3100 | -68.3650 | -21.1333 | -0.9548 | -0.8860 |
| 0.0 | 4.0 | 750 | 0.0000 | 1.8821 | -13.4603 | 1.0 | 15.3424 | -68.4222 | -21.1259 | -0.9551 | -0.8863 |
| 0.0 | 4.2667 | 800 | 0.0000 | 1.8839 | -13.4565 | 1.0 | 15.3404 | -68.4148 | -21.1223 | -0.9552 | -0.8862 |
| 0.0 | 4.5333 | 850 | 0.0000 | 1.8851 | -13.4374 | 1.0 | 15.3225 | -68.3765 | -21.1199 | -0.9552 | -0.8866 |
| 0.0 | 4.8 | 900 | 0.0000 | 1.8887 | -13.4633 | 1.0 | 15.3521 | -68.4284 | -21.1127 | -0.9547 | -0.8861 |
| 0.0 | 5.0667 | 950 | 0.0000 | 1.8879 | -13.4648 | 1.0 | 15.3527 | -68.4314 | -21.1144 | -0.9549 | -0.8862 |
| 0.0 | 5.3333 | 1000 | 0.0000 | 1.8884 | -13.4648 | 1.0 | 15.3532 | -68.4313 | -21.1134 | -0.9549 | -0.8862 |
Framework versions
- Transformers 4.44.2
- Pytorch 2.4.0+cu121
- Datasets 2.21.0
- Tokenizers 0.19.1
- Downloads last month
- 2
Inference Providers
NEW
This model is not currently available via any of the supported third-party Inference Providers, and
the model is not deployed on the HF Inference API.
Model tree for tsavage68/Na_L3_1000steps_1e6rate_05beta_cSFTDPO
Base model
meta-llama/Meta-Llama-3-8B-Instruct
Finetuned
tsavage68/Na_L3_100steps_1e6rate_SFT