README.md

metadata

license: apache-2.0
tags:
  - moe
  - frankenmoe
  - merge
  - mergekit
  - lazymergekit
  - yanolja/EEVE-Korean-10.8B-v1.0
  - upstage/SOLAR-10.7B-v1.0
base_model:
  - yanolja/EEVE-Korean-10.8B-v1.0
  - upstage/SOLAR-10.7B-v1.0

TEST_MODEL

TEST_MODEL is a Mixture of Experts (MoE) made with the following models using LazyMergekit:

• yanolja/EEVE-Korean-10.8B-v1.0
• upstage/SOLAR-10.7B-v1.0

🧩 Configuration

base_model: yanolja/EEVE-Korean-10.8B-v1.0
dtype: float16
experts:
  - source_model: yanolja/EEVE-Korean-10.8B-v1.0
    positive_prompts: ["You are an helpful general-pupose assistant."]
  - source_model: upstage/SOLAR-10.7B-v1.0
    positive_prompts: ["You are helpful assistant."]
merge_method: slerp
gate_mode: cheap_embed
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

tokenizer_source: base

💻 Usage

!pip install -qU transformers bitsandbytes accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "jieunhan/TEST_MODEL"

tokenizer = AutoTokenizer.from_pretrained(model)
pipeline = transformers.pipeline(
    "text-generation",
    model=model,
    model_kwargs={"torch_dtype": torch.float16, "load_in_4bit": True},
)

messages = [{"role": "user", "content": "Explain what a Mixture of Experts is in less than 100 words."}]
prompt = pipeline.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
print(outputs[0]["generated_text"])