nlile/PE-12b-pythia
014
<!-- 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. -->
PE-12b-pythia
This model is a fine-tuned version of EleutherAI/pythia-12b-deduped on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.1421
- Rewards/chosen: 3.5045
- Rewards/rejected: -2.3171
- Rewards/accuracies: 0.9441
- Rewards/margins: 5.8216
- Logps/rejected: -95.5639
- Logps/chosen: -116.1507
- Logits/rejected: -0.4604
- Logits/chosen: -0.4355
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: 3e-07
- trainbatchsize: 1
- evalbatchsize: 2
- seed: 42
- distributed_type: multi-GPU
- num_devices: 8
- gradientaccumulationsteps: 8
- totaltrainbatch_size: 64
- totalevalbatch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lrschedulertype: linear
- lrschedulerwarmup_ratio: 0.1
- num_epochs: 3
Training results
Framework versions
- Transformers 4.35.0
- Pytorch 2.1.1+cu121
- Datasets 2.14.6
- Tokenizers 0.14.1
