How to Convert YOLOv10 Model to TFLite with INT8 Quantization? about ultralytics HOT 2 OPEN

👋 Hello @AhmedFkih, thank you for your interest in Ultralytics YOLOv8 🚀! We recommend a visit to the Docs for new users where you can find many Python and CLI usage examples and where many of the most common questions may already be answered.

If this is a 🐛 Bug Report, please provide a minimum reproducible example to help us debug it.

If this is a custom training ❓ Question, please provide as much information as possible, including dataset image examples and training logs, and verify you are following our Tips for Best Training Results.

Join the vibrant Ultralytics Discord 🎧 community for real-time conversations and collaborations. This platform offers a perfect space to inquire, showcase your work, and connect with fellow Ultralytics users.

Install

Pip install the ultralytics package including all requirements in a Python>=3.8 environment with PyTorch>=1.8.

pip install ultralytics

Environments

YOLOv8 may be run in any of the following up-to-date verified environments (with all dependencies including CUDA/CUDNN, Python and PyTorch preinstalled):

• Notebooks with free GPU:
• Google Cloud Deep Learning VM. See GCP Quickstart Guide
• Amazon Deep Learning AMI. See AWS Quickstart Guide
• Docker Image. See Docker Quickstart Guide

Status

If this badge is green, all Ultralytics CI tests are currently passing. CI tests verify correct operation of all YOLOv8 Modes and Tasks on macOS, Windows, and Ubuntu every 24 hours and on every commit.

from ultralytics.

@AhmedFkih hello,

To convert a YOLOv10 model to TensorFlow Lite (TFLite) with INT8 quantization, you'll generally follow these steps:

1. Export YOLOv10 to ONNX or SavedModel: First, export your model to a compatible format like ONNX or TensorFlow's SavedModel.

2. Convert to TensorFlow Lite: Use TensorFlow's TFLite Converter to convert the model from ONNX/SavedModel to TFLite. During this step, you can specify the INT8 quantization.

3. Calibration: For INT8 quantization, you'll need to perform calibration using a representative dataset. This helps in accurately mapping the floating-point values to INT8.

Here’s a basic example using TensorFlow's TFLite Converter:

import tensorflow as tf

# Load the SavedModel
model = tf.saved_model.load('path_to_saved_model')

# Set up the converter with INT8 quantization
converter = tf.lite.TFLiteConverter.from_saved_model(saved_model_dir)
converter.optimizations = [tf.lite.Optimize.DEFAULT]
converter.representative_dataset = representative_data_gen
converter.target_spec.supported_ops = [tf.lite.OpsSet.TFLITE_BUILTINS_INT8]
converter.inference_input_type = tf.int8  # or tf.uint8
converter.inference_output_type = tf.int8  # or tf.uint8

# Convert the model
tflite_model = converter.convert()

# Save the TFLite model
with open('model.tflite', 'wb') as f:
    f.write(tflite_model)

For the representative_data_gen, you need to provide a function that yields batches of input data from your dataset. This data is used to calibrate the quantization parameters.

For more detailed guidance and advanced configurations, please refer to the TensorFlow Lite documentation on model optimization and quantization.

Best of luck with your deployment!

— The Ultralytics Team

from ultralytics.