metadata
license: apache-2.0
library_name: peft
tags:
- generated_from_trainer
base_model: mnoukhov/pythia410m-sft-tldr
model-index:
- name: pythia410m-dpo2-tldr
results: []
pythia410m-dpo2-tldr
This model is a fine-tuned version of mnoukhov/pythia410m-sft-tldr on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.6553
- Rewards/chosen: -0.0115
- Rewards/rejected: -0.0984
- Rewards/accuracies: 0.6726
- Rewards/margins: 0.0869
- Logps/rejected: -65.8898
- Logps/chosen: -65.8898
- Logps/ref Rejected: -59.5615
- Logps/ref Chosen: -65.6594
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: 1e-05
- train_batch_size: 4
- eval_batch_size: 8
- seed: 42
- distributed_type: multi-GPU
- num_devices: 4
- gradient_accumulation_steps: 4
- total_train_batch_size: 64
- total_eval_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- num_epochs: 1.0
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logps/ref Rejected | Logps/ref Chosen |
|---|---|---|---|---|---|---|---|---|---|---|---|
| No log | 0.2016 | 63 | 0.6719 | 0.0165 | -0.0292 | 0.6739 | 0.0457 | -65.3302 | -65.3302 | -59.5615 | -65.6594 |
| 0.6865 | 0.4032 | 126 | 0.6614 | 0.0030 | -0.0680 | 0.6752 | 0.0710 | -65.5994 | -65.5994 | -59.5615 | -65.6594 |
| 0.6865 | 0.6048 | 189 | 0.6571 | -0.0046 | -0.0866 | 0.6744 | 0.0820 | -65.7515 | -65.7515 | -59.5615 | -65.6594 |
| 0.6771 | 0.8064 | 252 | 0.6553 | -0.0115 | -0.0984 | 0.6726 | 0.0869 | -65.8898 | -65.8898 | -59.5615 | -65.6594 |
Framework versions
- PEFT 0.10.0
- Transformers 4.40.2
- Pytorch 2.2.1+cu121
- Datasets 2.19.1
- Tokenizers 0.19.1