---
license: apache-2.0
base_model: microsoft/phi-2
language:
- en
pipeline_tag: text-generation
tags:
- alignment-handbook
- generated_from_trainer
---

# outputs
This model is a fine-tuned version of [microsoft/phi-2](https://huggingface.co/microsoft/phi-2) using [SPIN](https://github.com/uclaml/SPIN) on [ultrachat_200k dataset](https://huggingface.co/datasets/HuggingFaceH4/ultrachat_200k).

# What's new
I think SPIN not only can use on a SFT model, but also it  can use on a pretrained model. 
Therefore, I use SPIN on a pretrained model microsoft/phi-2. And I get a higher score better than origin pretrained model. You can check the [open llm leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard).

The best paradigm for training a conversational Large Language Model (LLM):
pretrain -> dpo(spin) -> sft -> dpo(spin)

## Training procedure

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 5e-07
- train_batch_size: 1
- eval_batch_size: 1
- seed: 42
- distributed_type: multi-GPU
- num_devices: 8
- gradient_accumulation_steps: 8
- total_train_batch_size: 64
- total_eval_batch_size: 8
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 3

### Framework versions

- Transformers 4.37.0
- Pytorch 2.1.2+cu121
- Datasets 2.14.6
- Tokenizers 0.15.2