metadata
library_name: peft
tags:
- alignment-handbook
- trl
- dpo
- generated_from_trainer
base_model: NorLLM-AI/NorMistral-7B
datasets:
- hugodk-sch/aftonposten_title_prefs
model-index:
- name: norllm-ai-normistral-7b-align-scan
results: []
norllm-ai-normistral-7b-align-scan
This model is a fine-tuned version of data/norllm-ai-normistral-7b-sft-qlora on the hugodk-sch/aftonposten_title_prefs dataset. It achieves the following results on the evaluation set:
- Loss: 0.8067
- Rewards/chosen: -1.1692
- Rewards/rejected: -1.5184
- Rewards/accuracies: 0.5918
- Rewards/margins: 0.3492
- Logps/rejected: -37.2289
- Logps/chosen: -33.2312
- Logits/rejected: -2.8266
- Logits/chosen: -2.8291
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 | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.6746 | 0.26 | 100 | 0.6828 | 0.0185 | -0.0185 | 0.5694 | 0.0370 | -34.7290 | -31.2516 | -2.8058 | -2.8084 |
| 0.6195 | 0.52 | 200 | 0.6735 | -0.0458 | -0.1322 | 0.5511 | 0.0864 | -34.9185 | -31.3587 | -2.8176 | -2.8201 |
| 0.5567 | 0.78 | 300 | 0.6810 | -0.1233 | -0.2426 | 0.5723 | 0.1192 | -35.1024 | -31.4880 | -2.8203 | -2.8231 |
| 0.2251 | 1.04 | 400 | 0.6779 | -0.3249 | -0.4970 | 0.6013 | 0.1720 | -35.5264 | -31.8240 | -2.8175 | -2.8204 |
| 0.2082 | 1.3 | 500 | 0.6859 | -0.4136 | -0.6723 | 0.6092 | 0.2587 | -35.8186 | -31.9717 | -2.8475 | -2.8487 |
| 0.2119 | 1.56 | 600 | 0.6993 | -0.5421 | -0.7899 | 0.5926 | 0.2478 | -36.0147 | -32.1860 | -2.8301 | -2.8322 |
| 0.1579 | 1.82 | 700 | 0.7178 | -0.6062 | -0.8251 | 0.5806 | 0.2189 | -36.0734 | -32.2928 | -2.8261 | -2.8284 |
| 0.0649 | 2.08 | 800 | 0.7260 | -0.7190 | -1.0000 | 0.6071 | 0.2810 | -36.3648 | -32.4808 | -2.8243 | -2.8271 |
| 0.1014 | 2.34 | 900 | 0.7758 | -1.0050 | -1.3365 | 0.5831 | 0.3315 | -36.9256 | -32.9574 | -2.8278 | -2.8304 |
| 0.0425 | 2.6 | 1000 | 0.7952 | -1.0994 | -1.4459 | 0.5826 | 0.3465 | -37.1080 | -33.1148 | -2.8238 | -2.8267 |
| 0.0878 | 2.86 | 1100 | 0.7929 | -1.0931 | -1.4389 | 0.5889 | 0.3458 | -37.0962 | -33.1042 | -2.8257 | -2.8283 |
| 0.0534 | 3.12 | 1200 | 0.7997 | -1.1321 | -1.4857 | 0.5889 | 0.3535 | -37.1742 | -33.1693 | -2.8258 | -2.8285 |
| 0.035 | 3.38 | 1300 | 0.8024 | -1.1445 | -1.5019 | 0.5889 | 0.3575 | -37.2014 | -33.1899 | -2.8266 | -2.8291 |
| 0.0126 | 3.64 | 1400 | 0.8126 | -1.1630 | -1.5088 | 0.5860 | 0.3457 | -37.2128 | -33.2208 | -2.8267 | -2.8294 |
| 0.0525 | 3.9 | 1500 | 0.8088 | -1.1685 | -1.5136 | 0.5918 | 0.3451 | -37.2208 | -33.2299 | -2.8265 | -2.8292 |
Framework versions
- PEFT 0.10.0
- Transformers 4.39.0.dev0
- Pytorch 2.1.2+cu121
- Datasets 2.14.6
- Tokenizers 0.15.1