Emo-v1

A lightweight 3B parameter model fine-tuned for Reasoning.
Specialized in Algebra, Logic Puzzles, and Step-by-Step Reasoning.

📖 Model Description

Emo-v1 is a fine-tuned version of Qwen/Qwen2.5-3B-Instruct, optimized for mathematical reasoning and logic.

Unlike standard chat models that often guess answers, Emo-Qwen is trained to decompose problems into explicit steps before providing a final solution. It mimics the "Chain of Thought" (CoT) process found in larger reasoning models (like OpenAI's o1), making it surprisingly capable for its small size.

Key Features

Step-by-Step Reasoning: Forces a "Let's break this down" approach to minimize logic errors.
Math Specialist: Trained on the nvidia/OpenMathInstruct-2 dataset, covering algebra, calculus, and probability.
LaTeX Support: Optimized to output mathematical formulas in clean LaTeX format (e.g., $x^2 + y^2$).
Efficient: At only 3 Billion parameters, it runs on consumer hardware (even free Kaggle/Colab T4 GPUs) with low latency.

How to Use

Python Inference Code

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

# 1. Load Model
model_id = "PrimeTJ/Emo-v1"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.float16,
    device_map="auto"
)

# 2. Define the Prompt
system_prompt = "You are a helpful math assistant. Think step by step."
user_prompt = "A bat and a ball cost $1.10 in total. The bat costs $1.00 more than the ball. How much does the ball cost?"

messages = [
    {"role": "system", "content": system_prompt},
    {"role": "user", "content": user_prompt}
]

# 3. Generate
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)

generated_ids = model.generate(
    **model_inputs,
    max_new_tokens=1024,
    temperature=0.6  # Low temperature for logic
)

response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
print(response)