InvariantMind-Worker-7B (Worker Engineer - 7B)

InvariantMind-Worker-7B is the dedicated computational engineering and simulation synthesis engine of the dual-tier InvariantMind architecture. While the Tier 1 Oracle (InvariantMind-v1-14B) excels at high-level epistemic synthesis, hypothesis generation, and formal proofs, Tier 2 Worker-7B is specialized in transforming theoretical conjectures into verified, vectorized simulation code, numerical ODE/PDE integrators, and automated empirical experimentation.

Architecture & Lineage

• Base Model: Qwen/Qwen2.5-Coder-7B-Instruct
• Fine-Tuning Method: 4-bit NormalFloat QLoRA ($r=64, \alpha=128$)
• Colony Training Data: 2,330 curated scientific episodes (19.4 MB) encompassing rigorous simulation scripts, verification suites, and epistemic tool calls.
• Target Capabilities: High-performance vectorized scientific computing (NumPy, SciPy, Numba), stiff differential equation solvers, cellular automata engines, and conservation-law invariant checks.

How to Use

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
from peft import PeftModel

base_model_name = "Qwen/Qwen2.5-Coder-7B-Instruct"
adapter_id = "Ninitje/InvariantMind-Worker-7B"

bnb_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_compute_dtype=torch.bfloat16,
    bnb_4bit_quant_type="nf4"
)

tokenizer = AutoTokenizer.from_pretrained(base_model_name)
base_model = AutoModelForCausalLM.from_pretrained(
    base_model_name,
    quantization_config=bnb_config,
    device_map="auto"
)
model = PeftModel.from_pretrained(base_model, adapter_id)

prompt = "Write an optimized, vectorized Python implementation using NumPy to simulate N coupled Kuramoto oscillators and compute order parameter R(t)."
messages = [{"role": "user", "content": prompt}]
inputs = tokenizer(tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True), return_tensors="pt").to("cuda")

with torch.no_grad():
    outputs = model.generate(**inputs, max_new_tokens=600, temperature=0.6)
print(tokenizer.decode(outputs[0][inputs.input_ids.shape[1]:], skip_special_tokens=True))