clemsail/mascarade-esp32

Mascarade ESP32

Fine-tuned TinyLlama-1.1B-Chat model specialized in ESP32 microcontroller development.

Part of the Mascarade ecosystem — an agentic LLM orchestration system with domain-specific fine-tuned models for embedded systems and electronics.

Training details

Parameter	Value
Base model	TinyLlama/TinyLlama-1.1B-Chat-v1.0
Method	LoRA (PEFT) — merged into full weights
LoRA rank (r)	16
LoRA alpha	32
LoRA dropout	0.05
Target modules	q_proj, k_proj, v_proj, o_proj
Epochs	2
Training steps	30
Final train loss	1.3873
Dataset	ShareGPT format, domain-specific ESP32 examples
GPU	Quadro P2000 (5 GB VRAM)
Framework	Hugging Face Transformers + PEFT

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("electron-rare/mascarade-esp32")
tokenizer = AutoTokenizer.from_pretrained("electron-rare/mascarade-esp32")

messages = [{"role": "user", "content": "How do I configure deep sleep on ESP32-S3?"}]
inputs = tokenizer.apply_chat_template(messages, return_tensors="pt")
outputs = model.generate(inputs, max_new_tokens=512)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Related models

Model	Domain	Base
mascarade-iot	IoT general	Qwen2.5-Coder-1.5B
mascarade-spice	SPICE circuit simulation	TinyLlama-1.1B
mascarade-platformio	PlatformIO development	TinyLlama-1.1B

Datasets

All training datasets are available under clemsail on Hugging Face.