pico-type 0.1.5

A tiny byte-level multi-head content classifier

---

license: apache-2.0 language:

• en
• multilingual tags:
• byte-level
• content-classification
• onnx
• edge-ai
• matryoshka pipeline_tag: text-classification library_name: generic inference: parameters: provider: CPUExecutionProvider

---

pico-type 🔍

A tiny byte-level multi-head content classifier — ~1.5M params, ~209KB ONNX, <6ms inference.

Classifies any content into 7 categories from raw bytes: coarse type, modality, subtype, code language, text language, file MIME, and risk flags.

---

✨ Features

• No tokenizer — operates directly on raw UTF-8 bytes (supports all languages)
• 7 heads, one forward pass — coarse type, modality, subtype, code lang, text lang, file MIME, risk
• 4 Matryoshka tiers — tiny (16d) → small (64d) → base (192d) → pro (576d)
• ~200KB ONNX — deploy on edge devices, serverless, browser (WebAssembly)
• <12ms inference on CPU via ONNX Runtime
• CLI, Gradio Space, MCP server — ready to use

📊 Performance

Head	Classes	Accuracy
coarse	12	100%
modality	8	100%
subtype	24	98.8%
code_lang	62	61.3%
text_lang	30	100%
file_mime	90	100%
risk (mAP)	6	100%

1000 evaluation samples, base tier, ~5.6ms inference. Hindi (hi) added as a text language.

🚀 Quick Start

CLI

pip install picotype

echo "def hello():\n    return 42" | picotype --pretty
picotype --file document.txt
picotype --clip

Python

from model.pico_type.cli import load_onnx_model, run_onnx

session = load_onnx_model("base", "checkpoints")
result = run_onnx(session, "def hello(): pass")
print(result)

MCP Server (Claude/Cursor)

PICOTYPE_MODEL_DIR=./checkpoints python -m model.pico_type.mcp_server

🏗 Architecture

Bytes → ByteEmbed(256→96d) → 3×Conv1D(k=3,5,7) → 2×BiAttention(RoPE) → Pool(mean‖max‖std) → 7×Matryoshka Heads

• ByteEmbed: lookup-free byte embedding (256 vocab, 96 dim)
• Conv1D: 3 parallel kernels (width 3, 5, 7) with residual + layer norm
• BiAttention: bidirectional self-attention with RoPE, 4 heads, 96 dim
• Pool: mean + max + std concatenation
• Matryoshka Heads: 4 slices of the pooled vector (16/64/192/576 dim) → 7 linear classifiers

Total parameters: 1.43M (tiny) / 1.45M (small) / 1.48M (base) / 1.56M (pro)

🔧 Model Tiers

Tier	Dim	Params	ONNX Size
tiny	16	1.43M	207 KB
small	64	1.45M	207 KB
base	192	1.48M	209 KB
pro	576	1.56M	206 KB

All tiers share the same trunk; only the final linear layers differ.

🧪 Classification Heads

Head	Classes	Examples
coarse	12	text, code, link, image, file, config, markup, data, error, secret, archive, binary
modality	8	textual, binary_image, binary_archive, binary_executable, etc.
subtype	24	json, yaml, toml, csv, html, markdown, sql, log, dockerfile, etc.
code_lang	62	python, javascript, typescript, java, c, cpp, go, rust, etc.
text_lang	30	en, es, fr, de, it, pt, ru, zh, ja, ko, ar, hi, etc.
file_mime	90	text/html, application/json, application/pdf, image/png, video/mp4, etc.
risk	6	api_key, jwt, password, email, phone, ssh_key

🌐 Deployment

Platform	Location
HuggingFace Space	eulogik/pico-type
HuggingFace Model	eulogik/pico-type
GitHub	eulogik/pico-type
PyPI	pip install picotype

📚 Documentation

• Model Card — detailed architecture, training, and evaluation
• Architecture Plan — full design document
• Walkthrough — development log

📄 License

Apache 2.0