hugodk-sch's picture
End of training
e33c89a verified
|
raw
history blame
2.61 kB
metadata
library_name: peft
tags:
  - alignment-handbook
  - trl
  - dpo
  - generated_from_trainer
base_model: norallm/normistral-7b-warm
datasets:
  - hugodk-sch/aftonposten_title_prefs
model-index:
  - name: ap-normistral-7b-align-scan
    results: []

ap-normistral-7b-align-scan

This model is a fine-tuned version of data/ap-normistral-7b-sft-qlora on the hugodk-sch/aftonposten_title_prefs dataset. It achieves the following results on the evaluation set:

  • Loss: 0.7822
  • Rewards/chosen: -0.1005
  • Rewards/rejected: -0.2047
  • Rewards/accuracies: 0.5249
  • Rewards/margins: 0.1042
  • Logps/rejected: -36.2224
  • Logps/chosen: -32.5688
  • Logits/rejected: 98.7467
  • Logits/chosen: 98.7645

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
0.7017 0.26 100 0.8030 -0.0408 -0.0544 0.5137 0.0136 -36.0346 -32.4942 98.7637 98.7765
0.6236 0.52 200 0.7719 -0.1720 -0.2931 0.5216 0.1211 -36.3329 -32.6582 98.7039 98.7271
0.5655 0.78 300 0.7744 -0.0855 -0.2061 0.5428 0.1206 -36.2242 -32.5501 98.7517 98.7722

Framework versions

  • PEFT 0.10.0
  • Transformers 4.39.0.dev0
  • Pytorch 2.1.2+cu121
  • Datasets 2.14.6
  • Tokenizers 0.15.1