metadata
license: apache-2.0
library_name: peft
tags:
- alignment-handbook
- generated_from_trainer
- trl
- dpo
- generated_from_trainer
datasets:
- HuggingFaceH4/ultrafeedback_binarized
base_model: alignment-handbook/zephyr-7b-sft-full
model-index:
- name: zephyr-7b-dpo-lora-pubmedqa-selfgen-ultrafeedback2
results: []
zephyr-7b-dpo-lora-pubmedqa-selfgen-ultrafeedback2
This model is a fine-tuned version of EllieS/zephyr-7b-dpo-lora-pubmedqa-selfgen-complete on the HuggingFaceH4/ultrafeedback_binarized dataset. It achieves the following results on the evaluation set:
- Loss: 0.5869
- Rewards/chosen: -0.2480
- Rewards/rejected: -0.5522
- Rewards/accuracies: 0.7090
- Rewards/margins: 0.3042
- Logps/rejected: -306.5199
- Logps/chosen: -309.5233
- Logits/rejected: -2.5197
- Logits/chosen: -2.5501
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: 2
- eval_batch_size: 2
- seed: 42
- distributed_type: multi-GPU
- num_devices: 2
- gradient_accumulation_steps: 2
- total_train_batch_size: 8
- total_eval_batch_size: 4
- 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 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.5526 | 0.39 | 3000 | 0.6013 | -0.1508 | -0.4054 | 0.7040 | 0.2546 | -291.8378 | -299.8007 | -2.5651 | -2.5923 |
| 0.5814 | 0.79 | 6000 | 0.5867 | -0.2418 | -0.5459 | 0.7080 | 0.3040 | -305.8824 | -308.9029 | -2.5202 | -2.5505 |
Framework versions
- PEFT 0.7.1
- Transformers 4.36.2
- Pytorch 2.1.2+cu121
- Datasets 2.14.6
- Tokenizers 0.15.2