--- license: other library_name: transformers datasets: - HuggingFaceH4/ultrafeedback_binarized base_model: wandb/gemma-2b-zephyr-sft license_name: gemma-terms-of-use license_link: https://ai.google.dev/gemma/terms --- [

](https://wandb.ai/llm_surgery/gemma-zephyr) # Gemma 2B Zephyr DPO The [Zephyr](https://huggingface.co/HuggingFaceH4/zephyr-7b-beta) DPO recipe applied on top of SFT finetuned Gemma 2B ## Model description - **Model type:** A 8.5B parameter GPT-like model fine-tuned on a mix of publicly available, synthetic datasets. - **Language(s) (NLP):** Primarily English - **Finetuned from model:** [wandb/gemma-2b-zephyr-sft](https://huggingface.co/wandb/gemma-2b-zephyr-sft/) ## Recipe We trained using the DPO script in [alignment handbook recipe](https://github.com/huggingface/alignment-handbook/blob/main/scripts/run_dpo.py) and logging to W&B Visit the [W&B workspace here](https://wandb.ai/llm_surgery/gemma-zephyr?nw=nwusercapecape) ## License This model has the same license as the [original Gemma model collection](https://ai.google.dev/gemma/terms) ## Compute provided by [Lambda Labs](https://lambdalabs.com/) - 8xA100 80GB node around 13 hours of training