MT-SLM-7B

MT-SLM-7B is a mixture of experts model, merging four specialized models to create a well-rounded AI capable of handling diverse tasks. It excels in coding, mathematical problem-solving, storytelling, and general-purpose chat interactions. The merging process was performed using LazyMergekit.

🧩 Component Models

MT-SLM-7B integrates four expert models:

Mathematics Expert
Finetuned for mathematical reasoning and problem-solving.
Coding Expert
Finetuned for generating high-quality Python and general programming code.
Chat Expert
A general-purpose conversational AI for everyday interactions.
Storytelling Expert
Finetuned for generating creative and engaging stories.

The individual models contributing to this mixture are:

Chat Model:
mlabonne/AlphaMonarch-7B – A general-purpose model for most interactions.
Code Model:
beowolx/CodeNinja-1.0-OpenChat-7B – A highly capable coding model.
Math Model:
mlabonne/NeuralDaredevil-7B – Specialized in mathematical reasoning with strong MMLU and GMS8K scores.
Role-Play & Storytelling Model:
SanjiWatsuki/Kunoichi-DPO-v2-7B – Known for high-quality storytelling and role-playing (MT-Bench score of 8.51).

This model supports an 8k context window for extended interactions.

🛠️ Model Configuration

slices:
  - sources:
      - model: jaiyeshchahar/ChatingDeveloper-7B-slerp
        layer_range: [0, 32]
      - model: jaiyeshchahar/storywriter-mathematician
        layer_range: [0, 32]

merge_method: slerp
base_model: jaiyeshchahar/storywriter-mathematician
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
---
license: apache-2.0
base_model:
- jaiyeshchahar/ChatingDeveloper-7B-slerp
- jaiyeshchahar/storywriter-mathematician
tags:
- merge
- mergekit
- lazymergekit
- jaiyeshchahar/ChatingDeveloper-7B-slerp
- jaiyeshchahar/storywriter-mathematician
---

This model supports an 8k context window for extended interactions.

🚀 Usage

1. Install Dependencies

Install the required libraries using pip:

pip install -qU transformers accelerate

2. Load the Model and Generate Text

Below is an example Python script to load the model and generate text:

from transformers import AutoTokenizer
import transformers
import torch

# Specify the model name
model = "jaiyeshchahar/MT-SLM-7B"

# Define your conversation as a list of messages
messages = [{"role": "user", "content": "What is a large language model?"}]

# Initialize the tokenizer and prepare the prompt
tokenizer = AutoTokenizer.from_pretrained(model)
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)

# Set up the text generation pipeline
pipeline = transformers.pipeline(
    "text-generation",
    model=model,
    torch_dtype=torch.float16,
    device_map="auto",
)

# Generate text output
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"])

3. Example Use Cases

Article Explanation: Summarize and explain complex articles.
Coding Assistance: Generate, debug, and explain Python code.
Mathematical Problem Solving: Handle computations and logical reasoning.
Creative Storytelling: Craft engaging narratives and role-play scenarios.

🎯 Conclusion

MT-SLM-7B is a powerful, well-rounded assistant that leverages a mixture of expert models to deliver exceptional performance across various domains. Whether you need a reliable coding companion, a math tutor, or a creative storyteller, this model is designed to meet your needs. Try it out and experience the full range of its capabilities!

Happy generating! 🚀

jaiyeshchahar
/

MT-SLM-7B