--- language: - ja - en tags: - merge - mergekit - lazymergekit - Rakuten/RakutenAI-7B-chat - lightblue/karasu-7B-chat-plus-unleashed base_model: - Rakuten/RakutenAI-7B-chat - lightblue/karasu-7B-chat-plus-unleashed --- # 🍊 Neroli-Rak-Lig-slerp-7B Neroli-Rak-Lig-slerp-7B is a merge of the following models using [LazyMergekit](https://colab.research.google.com/drive/1obulZ1ROXHjYLn6PPZJwRR6GzgQogxxb?usp=sharing) of [Maxime Labonne](https://huggingface.co/mlabonne) powered by [MergeKit](https://github.com/arcee-ai/mergekit) of [Arcee AI](https://www.arcee.ai): * [Rakuten/RakutenAI-7B-chat](https://huggingface.co/Rakuten/RakutenAI-7B-chat) (base model) * [lightblue/karasu-7B-chat-plus-unleashed](https://huggingface.co/lightblue/karasu-7B-chat-plus-unleashed) ## 💻 Configuration ```yaml slices: - sources: - model: Rakuten/RakutenAI-7B-chat layer_range: [0, 32] - model: lightblue/karasu-7B-chat-plus-unleashed layer_range: [0, 32] merge_method: slerp base_model: Rakuten/RakutenAI-7B-chat parameters: t: - filter: self_attn value: [0, 0.5, 0.3, 0.7, 1] - filter: mlp value: [1, 0.5, 0.7, 0.3, 0] - value: 0.5 dtype: bfloat16 ``` ## 🤗 Usage for HuggingFace ```python from transformers import AutoTokenizer, AutoModelForCausalLM from transformers import pipeline import torch model_name = "AkimfromParis/Neroli-Rak-Lig-slerp-7B" tokenizer = AutoTokenizer.from_pretrained(model_name) model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.bfloat16) pipe = pipeline("text-generation", model=model, tokenizer=tokenizer, pad_token_id=tokenizer.eos_token_id) messages = [ {"role": "system","content": "あなたは誠実で優秀な日本人のアシスタントです。以下のトピックに関する詳細な情報を提供してください。"}, {"role": "user", "content": "大谷翔平選手は誰ですか?"}, ] print(pipe(messages, max_new_tokens=256)[0]['generated_text'][-1]) ``` # 🔖 Citation ``` @misc{goddard2024arcee, title={Arcee's MergeKit: A Toolkit for Merging Large Language Models}, author={Goddard, Charles and Siriwardhana, Shamane and Ehghaghi, Malikeh and Meyers, Luke and Karpukhin, Vlad and Benedict, Brian and McQuade, Mark and Solawetz, Jacob}, journal={arXiv preprint arXiv:2403.13257}, year={2024} } ``` arxiv.org/abs/2403.13257