caid-technologies

parti-base

Deploy Dedicated

README

License: apache-2.0

What it can do

Give it a hardware idea and it can produce any of:

📋 a parts list (components)
🔌 a wiring/connection map between the parts
🛠️ ordered build steps
💲 rough sourcing and cost info
✅ a basic design check
📦 or the whole project plan at once

You can ask for the complete plan, or just one piece (like only the parts list).

What it's good for — and not

✅ Good for: brainstorming hardware projects, drafting parts lists and build steps, and turning a rough idea into an organized starting plan.

🚫 Not for: final engineering decisions, real CAD models, electrical safety, or anything safety-critical. Treat the output as a helpful first draft to review, not a finished design.

Try it

python
from transformers import AutoModelForCausalLM, AutoTokenizer

REPO = "caid-technologies/blueprint-base"
model = AutoModelForCausalLM.from_pretrained(REPO, device_map="auto", torch_dtype="bfloat16")
tok = AutoTokenizer.from_pretrained(REPO)

msgs = [
    {"role": "system", "content":
        "You design hobbyist electronics projects. Given a request, reply with a single "
        "JSON object describing the full project. Output only the JSON."},
    {"role": "user", "content": "A compact desk clock with an e-ink display and an IR remote."},
]
inputs = tok.apply_chat_template(
    msgs, add_generation_prompt=True, return_tensors="pt", return_dict=True).to(model.device)
out = model.generate(**inputs, max_new_tokens=6144, do_sample=False,
                     repetition_penalty=1.1, pad_token_id=tok.eos_token_id)
print(tok.decode(out[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True))

💡 Tip: keep do_sample=False (greedy decoding — sampling degrades the JSON output), keep max_new_tokens high (≥ 6000) so long plans aren't cut off, and keep repetition_penalty=1.1 so wiring lists don't get stuck repeating. For Ollama/local apps, convert this model to GGUF with llama.cpp.

What it learned from

It was trained on about 130 hobbyist hardware projects — things like weather stations, small robots, drones, smart-home gadgets, lab tools, and audio gear — expanded into a few thousand practice examples. Everything is small, maker-style electronics-plus-hardware.

Most common project types in the training data:

Table with columns: Project type, Share, Examples
Project type	Share	Examples
Test & lab instruments	~20%	function generator, Geiger counter
Smart-home / IoT gadgets	~15%	pet feeder, smart mailbox, pill dispenser
Radio, comms & networking	~9%	LoRa base station, APRS tracker, NAS
Wearables & health	~8%	sleep ring, heart-rate strap
Audio & music	~8%	synth module, guitar pedal, speaker
Robotics & motion

Good to know (limitations)

It's a small model, so complex, many-part projects are harder for it.
It proposes designs; it doesn't verify them. Always sanity-check before building.
It's strongest on common project types (lab tools, smart-home) and weaker on rarer ones (games, automotive).

How well it works

We tested it on projects it had never seen during training. Here's how often it produced a valid, well-structured result for each task:

Table with columns: Task, Valid result
Task	Valid result
🛠️ Build steps	~100%
✅ Design check	~100%
📋 Parts list	~95%
📦 Full project plan	~85–97%
🔌 Wiring map	~67%

It's strongest at build steps, design checks, and parts lists. Full end-to-end plans are close behind, and wiring maps are the hardest (and most sensitive to the repetition_penalty tip above). Figures are from held-out testing and are being finalized for the current version.

Base model: Qwen/Qwen2.5-3B-Instruct; this repo is the fine-tune merged to 16-bit (standalone, no adapter needed).
Method: QLoRA with Unsloth (LoRA r=32, alpha=32, all attention+MLP projections), then merged.
Training: 1 epoch, max_seq_len 6144, effective batch 8, lr 2e-4 (linear, 3% warmup), adamw_8bit, NEFTune α=5, loss masked to assistant turns, early stopping on eval loss
Hardware: single RTX 4070 (12 GB)
Data: synthetic dataset projected into 6 task "modes" (full plan, parts, wiring, instructions, validation); split grouped by project so none leak between train/test. ~3,242 rows; modes rebalanced (cap 350/mode) so the model doesn't coast on the easy ones.
Inference: do_sample=False, repetition_penalty≈1.1, max_new_tokens≥6000, pass the attention mask.

bibtex
@misc{blueprint_base,
  title  = {Blueprint Base: Qwen2.5-3B for structured hardware project generation},
  author = {Caid Technologies},
  year   = {2026},
  howpublished = {\url{https://huggingface.co/caid-technologies}}
}

Available on FriendliAI

Dedicated Endpoints

Run this model inference on single tenant GPU with unmatched speed and reliability at scale.

Learn more

Container

Run this model inference with full control and performance in your environment.

Learn more

Model Details

Model Provider

caid-technologies

Model Tree

Base

Qwen/Qwen2.5-3B-Instruct

Fine-tuned

this model

Input Modalities

Text

Output Modalities