README.md

metadata

language:
  - ja
  - en
tags:
  - merge
  - mergekit
  - lazymergekit
  - elyza/ELYZA-japanese-Llama-2-7b
  - tokyotech-llm/Swallow-7b-hf
base_model:
  - elyza/ELYZA-japanese-Llama-2-7b
  - tokyotech-llm/Swallow-7b-hf

🌿 Heliotrope-Ely-Swa-slerp-7B

Heliotrope-Ely-Swa-slerp-7B is a merge of the following models using LazyMergekit of Maxime Labonne powered by MergeKit of Arcee AI:

• elyza/ELYZA-japanese-Llama-2-7b (Base model)
• tokyotech-llm/Swallow-7b-hf

💻 Configuration

slices:
  - sources:
      - model: elyza/ELYZA-japanese-Llama-2-7b
        layer_range: [0, 32]
      - model: tokyotech-llm/Swallow-7b-hf
        layer_range: [0, 32]
merge_method: slerp
base_model: elyza/ELYZA-japanese-Llama-2-7b
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

# !pip install -qU transformers accelerate

from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
import torch

model_name = "AkimfromParis/Heliotrope-Ely-Swa-slerp-7B"

tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

pipe = pipeline("text-generation",model=model, tokenizer=tokenizer, torch_dtype=torch.float16, device_map="auto")
sequences = pipe('大谷翔平選手は', do_sample=False, max_new_tokens=100)

print(sequences[0].get("generated_text"))

🔖 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