ONNXNPU Toolkit

ONNXNPU Toolkit is an open-source, lightweight CLI utility for ONNX model compatibility validation and performance optimization on Kneron NPUs (KL520–KL730). Built for machine learning engineers and edge-AI developers, it enables you to:

• Automatically detect unsupported ONNX operators before deployment
• Generate detailed JSON or Markdown reports for hardware-specific compatibility
• Customize hardware profiles and override rules to match your NPU target

Streamline your ONNX to NPU workflow, catch integration issues early, and boost edge-AI inference performance.

"Catch unsupported operators early before they derail your model." — Mason Huang

✨ Features (v0.1 — released)

🚀 Quick start

You can use two different CLI commands: onnxnpu or onpu to check, optimize, and modify your ONNX models for NPU deployment. Both commands provide identical functionality with the same syntax.

Sample output

════════════════════════════════════════════════════════════
MODEL INFO
════════════════════════════════════════════════════════════
Model name - my_model.onnx
IR version : 6
Opset : 13
Inputs  : input  float32  [1, 3, 112, 112]
Outputs : output  float32  [1, 512]
Dynamic axes : None detected ✓

════════════════════════════════════════════════════════════
HARDWARE COMPATIBILITY - KL520
════════════════════════════════════════════════════════════
+--------+--------------------+-------+-------+
| Status | Operator           | Count | Notes |
+--------+--------------------+-------+-------+
|   ✓    | Add                | 16    |       |
|   ✓    | BatchNormalization | 18    |       |
|   ✓    | Conv               | 37    |       |
|   ✓    | Flatten            | 1     |       |
|   ✓    | Gemm               | 1     |       |
|   ✓    | PRelu              | 17    |       |
+--------+--------------------+-------+-------+

════════════════════════════════════════════════════════════
MEMORY REQUIREMENTS - KL520
════════════════════════════════════════════════════════════
Estimated NEF size:   32.56 MB
USB model limit:      35.00 MB  -> OK
Flash model limit:    32.00 MB  -> MIGHT EXCEEDS LIMIT

Summary: All operators are supported on KL520 ✓
Total operators: 6 (instances: 90)

🧑‍💻 API usage

from onnxnpu import Checker, load_profile

checker = Checker("my_model.onnx", profile=load_profile("kl720"))
report = checker.run()
print(report.to_markdown())

# Update opset version
from onnxnpu import update_opset_version
update_opset_version("my_model.onnx", target_version=13)

📖 Hardware profiles

Profiles live under onnxnpu/profiles/*.json. Each profile declares the operators, attributes, and constraints supported by a particular accelerator. See docs/PROFILE_SCHEMA.md for the JSON schema.

Contributions for new hardware are very welcome!

A note on language

The primary language of this README is English for wider community reach. A Traditional Chinese translation will be added soon.