hugodk-sch's picture
End of training
6d38632 verified
|
raw
history blame
2.6 kB
metadata
library_name: peft
tags:
  - alignment-handbook
  - trl
  - dpo
  - generated_from_trainer
base_model: NbAiLab/nb-gpt-j-6B-v2
datasets:
  - hugodk-sch/aftonposten_title_prefs
model-index:
  - name: aftonposten-6b-align-scan
    results: []

aftonposten-6b-align-scan

This model is a fine-tuned version of data/ap-gpt-j-6b-sft-qlora-04-08 on the hugodk-sch/aftonposten_title_prefs dataset. It achieves the following results on the evaluation set:

  • Loss: 0.9479
  • Rewards/chosen: -0.0348
  • Rewards/rejected: -0.0895
  • Rewards/accuracies: 0.5419
  • Rewards/margins: 0.0548
  • Logps/rejected: -37.6161
  • Logps/chosen: -34.0732
  • Logits/rejected: -2.2190
  • Logits/chosen: -2.2239

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: 1

Training results

Training Loss Epoch Step Validation Loss Rewards/chosen Rewards/rejected Rewards/accuracies Rewards/margins Logps/rejected Logps/chosen Logits/rejected Logits/chosen
0.8913 0.26 100 0.9845 -0.0055 -0.0215 0.5195 0.0159 -37.5405 -34.0407 -2.2273 -2.2322
0.7293 0.52 200 0.9602 -0.0172 -0.0580 0.5714 0.0408 -37.5811 -34.0537 -2.2238 -2.2286
0.6144 0.78 300 0.9713 -0.0468 -0.0779 0.5282 0.0310 -37.6032 -34.0866 -2.2201 -2.2249

Framework versions

  • PEFT 0.8.2
  • Transformers 4.37.2
  • Pytorch 2.1.2+cu121
  • Datasets 2.17.0
  • Tokenizers 0.15.1