|
--- |
|
library_name: peft |
|
tags: |
|
- trl |
|
- dpo |
|
- alignment-handbook |
|
- generated_from_trainer |
|
base_model: NbAiLab/nb-gpt-j-6B-v2 |
|
model-index: |
|
- name: aftonposten-6b-align-scan |
|
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. --> |
|
|
|
# aftonposten-6b-align-scan |
|
|
|
This model is a fine-tuned version of [NbAiLab/nb-gpt-j-6B-v2](https://huggingface.co./NbAiLab/nb-gpt-j-6B-v2) on the None dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 1.6201 |
|
- Rewards/chosen: 0.0327 |
|
- Rewards/rejected: 0.0149 |
|
- Rewards/accuracies: 0.5249 |
|
- Rewards/margins: 0.0178 |
|
- Logps/rejected: -37.4793 |
|
- Logps/chosen: -33.9527 |
|
- Logits/rejected: -2.2332 |
|
- Logits/chosen: -2.2381 |
|
|
|
## 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: 2 |
|
- total_train_batch_size: 8 |
|
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
|
- lr_scheduler_type: cosine |
|
- lr_scheduler_warmup_ratio: 0.1 |
|
- num_epochs: 4 |
|
|
|
### Training results |
|
|
|
| Training Loss | Epoch | Step | Logits/chosen | Logits/rejected | Logps/chosen | Logps/rejected | Validation Loss | Rewards/accuracies | Rewards/chosen | Rewards/margins | Rewards/rejected | |
|
|:-------------:|:-----:|:----:|:-------------:|:---------------:|:------------:|:--------------:|:---------------:|:------------------:|:--------------:|:---------------:|:----------------:| |
|
| 1.4583 | 0.26 | 100 | -2.2357 | -2.2308 | -34.0303 | -37.5236 | 1.6162 | 0.5245 | 0.0017 | 0.0045 | -0.0028 | |
|
| 1.279 | 0.52 | 200 | -2.2359 | -2.2311 | -34.0825 | -37.5923 | 1.6100 | 0.5257 | -0.0192 | 0.0111 | -0.0303 | |
|
| 1.0043 | 0.78 | 300 | -2.2312 | -2.2263 | -34.0845 | -37.6004 | 1.5962 | 0.5166 | -0.0200 | 0.0135 | -0.0335 | |
|
| 0.7239 | 1.04 | 400 | 1.6461 | -0.0219 | -0.0311 | 0.5341 | 0.0092 | -37.5945 | -34.0893 | -2.2276 | -2.2324 | |
|
| 0.6061 | 1.3 | 500 | 1.6487 | -0.0274 | -0.0429 | 0.5395 | 0.0155 | -37.6239 | -34.1030 | -2.2282 | -2.2330 | |
|
| 0.9255 | 1.56 | 600 | 1.5912 | 0.0108 | -0.0119 | 0.5544 | 0.0228 | -37.5464 | -34.0074 | -2.2273 | -2.2321 | |
|
| 0.8252 | 1.82 | 700 | 1.6334 | 0.0226 | 0.0045 | 0.5216 | 0.0180 | -37.5053 | -33.9781 | -2.2298 | -2.2346 | |
|
| 0.2848 | 2.08 | 800 | 1.6033 | 0.0153 | -0.0031 | 0.5249 | 0.0184 | -37.5244 | -33.9964 | -2.2313 | -2.2361 | |
|
| 0.3671 | 2.34 | 900 | 1.6569 | 0.0283 | 0.0177 | 0.5162 | 0.0106 | -37.4723 | -33.9637 | -2.2309 | -2.2358 | |
|
| 0.3936 | 2.6 | 1000 | 1.6203 | 0.0348 | 0.0187 | 0.5428 | 0.0161 | -37.4698 | -33.9475 | -2.2325 | -2.2374 | |
|
| 0.3156 | 2.86 | 1100 | 1.6012 | 0.0302 | 0.0108 | 0.5606 | 0.0194 | -37.4896 | -33.9592 | -2.2326 | -2.2375 | |
|
| 0.2893 | 3.12 | 1200 | 1.5705 | 0.0346 | 0.0103 | 0.5365 | 0.0243 | -37.4909 | -33.9480 | -2.2335 | -2.2383 | |
|
| 0.277 | 3.38 | 1300 | 1.6102 | 0.0314 | 0.0121 | 0.5403 | 0.0194 | -37.4865 | -33.9559 | -2.2333 | -2.2382 | |
|
| 0.139 | 3.64 | 1400 | 1.6181 | 0.0273 | 0.0092 | 0.5307 | 0.0181 | -37.4937 | -33.9663 | -2.2333 | -2.2381 | |
|
| 0.24 | 3.9 | 1500 | 1.6201 | 0.0327 | 0.0149 | 0.5249 | 0.0178 | -37.4793 | -33.9527 | -2.2332 | -2.2381 | |
|
|
|
|
|
### Framework versions |
|
|
|
- PEFT 0.10.0 |
|
- Transformers 4.39.0.dev0 |
|
- Pytorch 2.1.2+cu121 |
|
- Datasets 2.14.6 |
|
- Tokenizers 0.15.1 |
