metadata

license: apache-2.0
tags:
  - moe
  - frankenmoe
  - merge
  - mergekit
  - lazymergekit
  - Locutusque/TinyMistral-248M-v2
  - Locutusque/TinyMistral-248M-v2.5
  - Locutusque/TinyMistral-248M-v2.5-Instruct
  - jtatman/tinymistral-v2-pycoder-instruct-248m
  - Felladrin/TinyMistral-248M-SFT-v4
  - Locutusque/TinyMistral-248M-v2-Instruct
base_model:
  - Locutusque/TinyMistral-248M-v2
  - Locutusque/TinyMistral-248M-v2.5
  - Locutusque/TinyMistral-248M-v2.5-Instruct
  - jtatman/tinymistral-v2-pycoder-instruct-248m
  - Felladrin/TinyMistral-248M-SFT-v4
  - Locutusque/TinyMistral-248M-v2-Instruct

TinyMistral-6x248M

TinyMistral-6x248M is a Mixure of Experts (MoE) made with the following models using LazyMergekit:

🧩 Configuration

base_model: Locutusque/TinyMistral-248M-v2.5
experts:
  - source_model: Locutusque/TinyMistral-248M-v2
    positive_prompts:
    - "An emerging"
    - "assistant"
    - "TITLE"
    - "begin"
  - source_model: Locutusque/TinyMistral-248M-v2.5
    positive_prompts:
    - "Python"
    - "C++"
    - "AI"
    - "textbook"
  - source_model: Locutusque/TinyMistral-248M-v2.5-Instruct
    positive_prompts:
    - "chemistry"
    - "biology"
    - "physics"
    - "math"
    - "history"
    - "code"
  - source_model: jtatman/tinymistral-v2-pycoder-instruct-248m
    positive_prompts:
    - "code"
    - "python"
    - "programming"
    - "algorithm"
  - source_model: Felladrin/TinyMistral-248M-SFT-v4
    positive_prompts:
    - "Escreba"
    - "Voici"
    - "Para"
    - "Cuales"
    - "Welche"
    - "If you had to imagine"
  - source_model: Locutusque/TinyMistral-248M-v2-Instruct
    positive_prompts:
    - "Write an essay"
    - "What are"
    - "instruct"
    - "How does"
    - "Identify the"

💻 Usage

!pip install -qU transformers bitsandbytes accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "M4-ai/TinyMistral-6x248M"

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"])