quantize.py

from functools import partial

from optimum.amd.ryzenai import (
    AutoQuantizationConfig,
    RyzenAIOnnxQuantizer,
)
from optimum.exporters.onnx import main_export
from transformers import AutoFeatureExtractor


# Define paths for exporting ONNX model and saving quantized model
export_dir = "resnet_onnx"
quantization_dir = "resnet_onnx_quantized"

# Specify the model ID from Transformers
model_id = "microsoft/resnet-18"

# Step 1: Export the model to ONNX format using Optimum Exporters
main_export(
    model_name_or_path=model_id,
    output=export_dir,
    task="image-classification",
    opset=13,
    batch_size=1,
    height=224,
    width=224,
    no_dynamic_axes=True,
)

# Step 2: Preprocess configuration and data transformations
feature_extractor = AutoFeatureExtractor.from_pretrained(model_id)


def preprocess_fn(ex, feature_extractor):
    image = ex["image"]
    if image.mode == "L":
        image = image.convert("RGB")
    pixel_values = feature_extractor(image).pixel_values[0]
    return {"pixel_values": pixel_values}


# Step 3: Initialize the RyzenAIOnnxQuantizer with the exported model
quantizer = RyzenAIOnnxQuantizer.from_pretrained(export_dir)

# Step 4: Load recommended quantization config for model
quantization_config = AutoQuantizationConfig.ipu_cnn_config()

# Step 5: Obtain a calibration dataset for computing quantization parameters
train_calibration_dataset = quantizer.get_calibration_dataset(
    "imagenet-1k",
    preprocess_function=partial(preprocess_fn, feature_extractor=feature_extractor),
    num_samples=100,
    dataset_split="train",
    preprocess_batch=False,
    streaming=True,
)

# Step 6: Run the quantizer with the specified configuration and calibration data
quantizer.quantize(
    quantization_config=quantization_config, dataset=train_calibration_dataset, save_dir=quantization_dir
)