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_05beta_cSFTDPO
results: []
IE_L3_1000steps_1e7rate_05beta_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: 0.1802
- Rewards/chosen: -1.1386
- Rewards/rejected: -10.9339
- Rewards/accuracies: 0.7400
- Rewards/margins: 9.7954
- Logps/rejected: -97.4951
- Logps/chosen: -85.0749
- Logits/rejected: -0.7939
- Logits/chosen: -0.7200
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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.4416 | 0.4 | 50 | 0.3457 | -0.0969 | -1.1506 | 0.7400 | 1.0537 | -77.9284 | -82.9916 | -0.7954 | -0.7373 |
| 0.1388 | 0.8 | 100 | 0.1803 | -0.7835 | -7.7662 | 0.7400 | 6.9827 | -91.1596 | -84.3647 | -0.7936 | -0.7251 |
| 0.1387 | 1.2 | 150 | 0.1802 | -0.9415 | -9.2178 | 0.7400 | 8.2763 | -94.0629 | -84.6808 | -0.7940 | -0.7226 |
| 0.1733 | 1.6 | 200 | 0.1802 | -0.9618 | -9.5890 | 0.7400 | 8.6272 | -94.8052 | -84.7213 | -0.7940 | -0.7227 |
| 0.2253 | 2.0 | 250 | 0.1802 | -1.0365 | -9.8116 | 0.7400 | 8.7750 | -95.2504 | -84.8709 | -0.7938 | -0.7219 |
| 0.1386 | 2.4 | 300 | 0.1802 | -1.0393 | -10.0428 | 0.7400 | 9.0035 | -95.7128 | -84.8764 | -0.7938 | -0.7216 |
| 0.1213 | 2.8 | 350 | 0.1802 | -1.0477 | -10.3216 | 0.7400 | 9.2739 | -96.2705 | -84.8933 | -0.7934 | -0.7207 |
| 0.1906 | 3.2 | 400 | 0.1802 | -1.0921 | -10.5149 | 0.7400 | 9.4228 | -96.6571 | -84.9820 | -0.7947 | -0.7217 |
| 0.1906 | 3.6 | 450 | 0.1802 | -1.0970 | -10.5317 | 0.7400 | 9.4347 | -96.6906 | -84.9917 | -0.7945 | -0.7214 |
| 0.208 | 4.0 | 500 | 0.1802 | -1.1136 | -10.7153 | 0.7400 | 9.6017 | -97.0578 | -85.0249 | -0.7951 | -0.7219 |
| 0.156 | 4.4 | 550 | 0.1802 | -1.1237 | -10.8074 | 0.7400 | 9.6837 | -97.2419 | -85.0451 | -0.7948 | -0.7214 |
| 0.1213 | 4.8 | 600 | 0.1802 | -1.1291 | -10.8336 | 0.7400 | 9.7045 | -97.2944 | -85.0559 | -0.7943 | -0.7205 |
| 0.1906 | 5.2 | 650 | 0.1802 | -1.1297 | -10.8980 | 0.7400 | 9.7683 | -97.4233 | -85.0572 | -0.7939 | -0.7202 |
| 0.2426 | 5.6 | 700 | 0.1802 | -1.1277 | -10.8859 | 0.7400 | 9.7582 | -97.3990 | -85.0531 | -0.7953 | -0.7215 |
| 0.2599 | 6.0 | 750 | 0.1802 | -1.1398 | -10.9204 | 0.7400 | 9.7806 | -97.4681 | -85.0774 | -0.7944 | -0.7204 |
| 0.1213 | 6.4 | 800 | 0.1802 | -1.1496 | -10.9309 | 0.7400 | 9.7813 | -97.4891 | -85.0970 | -0.7947 | -0.7207 |
| 0.2426 | 6.8 | 850 | 0.1802 | -1.1208 | -10.9075 | 0.7400 | 9.7867 | -97.4422 | -85.0394 | -0.7944 | -0.7204 |
| 0.1733 | 7.2 | 900 | 0.1802 | -1.1302 | -10.9173 | 0.7400 | 9.7871 | -97.4618 | -85.0581 | -0.7939 | -0.7201 |
| 0.1386 | 7.6 | 950 | 0.1802 | -1.1386 | -10.9339 | 0.7400 | 9.7954 | -97.4951 | -85.0749 | -0.7939 | -0.7200 |
| 0.156 | 8.0 | 1000 | 0.1802 | -1.1386 | -10.9339 | 0.7400 | 9.7954 | -97.4951 | -85.0749 | -0.7939 | -0.7200 |
Framework versions
- Transformers 4.44.2
- Pytorch 2.0.0+cu117
- Datasets 3.0.0
- Tokenizers 0.19.1