README.md

metadata

license: apache-2.0
tags:
  - moe
  - frankenmoe
  - merge
  - mergekit
  - lazymergekit
  - MaziyarPanahi/Calme-7B-Instruct-v0.9
  - S-miguel/The-Trinity-Coder-7B
  - grimjim/kukulemon-7B
base_model:
  - MaziyarPanahi/Calme-7B-Instruct-v0.9
  - S-miguel/The-Trinity-Coder-7B
  - grimjim/kukulemon-7B

Lumina-5-Instruct

Lumina-5-Instruct is a Mixture of Experts (MoE) made with the following models using LazyMergekit:

• MaziyarPanahi/Calme-7B-Instruct-v0.9
• S-miguel/The-Trinity-Coder-7B
• grimjim/kukulemon-7B

🧩 Configuration

base_model: MaziyarPanahi/Calme-7B-Instruct-v0.9
experts:
  - source_model: MaziyarPanahi/Calme-7B-Instruct-v0.9
    positive_prompts:
    - "chat"
    - "assistant"
  - source_model: S-miguel/The-Trinity-Coder-7B
    positive_prompts:
    - "code"
    - "reason"
    - "math"
  - source_model: grimjim/kukulemon-7B
    positive_prompts:
    - "roleplay"
    - "write"
    - "scene"
    - "story"

💻 Usage

!pip install -qU transformers bitsandbytes accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "Ppoyaa/Lumina-5-Instruct"

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"])