Nexus Swarm Worker — SFT Adapter (Qwen3-8B)
LoRA adapter for the worker node in the NEXUS Swarm v4a-max distributed offensive security agent system. Fine-tuned from Qwen/Qwen3-8B via supervised fine-tuning on role-specific tool-use data.
Role: Worker
The worker is the executor of the swarm. It receives specific, scoped subtasks from the orchestrator and executes them immediately with minimal reasoning. Designed to run on an RTX 3070 Laptop (8 GB) on the control node.
Core behavior:
- Execute delegated tasks with correct tool calls — no re-planning
- Keep reasoning to one sentence maximum before each tool call
- Parse and structure tool output for the orchestrator to consume
- Never delegate to other nodes — the worker is the executor
- Thinking mode disabled for fast execution
Pipeline Position
DAPT (offsec corpus) → SFT (this model) → FARCA-GRPO → DPO → GGUF
| Stage | Model |
|---|---|
| Base | Qwen/Qwen3-8B |
| SFT (this) | atlas-institute/qwen3-8b-worker-sft |
| FARCA-GRPO | atlas-institute/qwen3-8b-worker-farca |
| GGUF | atlas-institute/qwen3-8b-worker-gguf |
Swarm Topology
| Node | GPU | Role | Model |
|---|---|---|---|
| nexus-k1 | RTX 5060 Ti 16 GB | Orchestrator | Qwen3-14B Q4_K_M |
| nexus-control | RTX 3070 8 GB | Worker | Qwen3-8B Q4_K_M |
| nexus-3060 | RTX 3060 6 GB | Explore | Gemma-3-4B-it Q8_0 |
| nexus-opi5p | RK3588 NPU | Triage | Qwen3.5-0.8B w8a8 |
| nexus-opi5b | RK3588S ARM | Embeddings | — |
Training Details
- Dataset: atlas-institute/code-trainer-v11-worker (12K rows)
- Epochs: 2
- Max sequence length: 4096
- LoRA rank / alpha: 32 / 64
- LoRA dropout: 0.05
- Learning rate: 1e-4
- Batch size: 1 (gradient accumulation 16, effective batch 16)
- Precision: BF16
- Hardware: NVIDIA A100 80 GB (HuggingFace Jobs)
- Thinking mode: Disabled (
enable_thinking: false)
Validation gate: tool_call_pass_rate ≥ 0.80 across 14 scenarios.
Usage
from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel
base_model = AutoModelForCausalLM.from_pretrained(
"Qwen/Qwen3-8B", torch_dtype="auto", device_map="auto"
)
model = PeftModel.from_pretrained(base_model, "atlas-institute/qwen3-8b-worker-sft")
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen3-8B")
messages = [
{"role": "system", "content": "You are a worker node. Execute the delegated task immediately."},
{"role": "user", "content": "Run nmap -sV -p 80,443 on 10.10.10.5 and return structured results."},
]
inputs = tokenizer.apply_chat_template(messages, return_tensors="pt").to(model.device)
outputs = model.generate(inputs, max_new_tokens=768)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
Developed By
Atlas Institute — Code-Trainer (RTPI) pipeline.
Framework Versions
- PEFT: 0.19.1
- Transformers: 5.7.0
- TRL: 1.3.0
- PyTorch: 2.11.0+cu128