Quark ONNX Quantization Tutorial For Resnet50#
This topic outlines best practice for Post-Training Quantization (PTQ) in Quark ONNX. It provides guidance on fine-tuning your quantization strategy to meet target quantization accuracy.
The example has the following parts:
Install requirements
Prepare model
Prepare data
Quantizatize model With different configs
XINT8
A8W8
A16W8
BF16
BFP16
CLE
AdaRound
AdaQuant
Exclude Nodes
Evaluate model
1) Install The Necessary Python Packages:#
In addition to Quark that must be installed as documented at here, extra packages are require for this tutorial.
%pip install torch torchvision --index-url https://download.pytorch.org/whl/cpu
%pip install amd-quark
%pip install -r ./requirements.txt
2) Download resnet50-v1-12 Model#
The model is publicly available and can be downloaded from this link: onnx/models
!mkdir -p models
!wget -O models/resnet50-v1-12.onnx https://github.com/onnx/models/raw/new-models/vision/classification/resnet/model/resnet50-v1-12.onnx
3) Prepare data#
ILSVRC 2012, commonly known as ‘ImageNet’. This dataset provides access to ImageNet (ILSVRC) 2012 which is the most commonly used subset of ImageNet. This dataset spans 1000 object classes and contains 50,000 validation images.
If you already have an ImageNet datasets, you can directly use your dataset path.
To prepare the test data, please check the download section of the main website: https://huggingface.co/datasets/imagenet-1k/tree/main/data. You need to register and download val_images.tar.gz to the current directory.
Then, create a val_data folder and decompress the .gz file to the folder.
!mkdir -p val_data && tar -xzf val_images.tar.gz -C val_data
If you have a local cache to store the dataset, you can use and
environment variable like LOCAL_DATA_CACHE to specify its path. This
is useful to organize and store all your datasets for different
experiments in a central place. Otherwise, the current folder is used,
and validation dataset and calibration dataset will be created under
current directory.
import os
import shutil
import sys
import time
source_folder = "val_data"
calib_data_path = "calib_data"
if os.environ.get("LOCAL_DATA_CACHE") is not None:
data_path = os.environ["LOCAL_DATA_CACHE"]
source_folder = os.path.join(data_path, "Imagenet/val")
calib_data_path = os.path.join(data_path, "Imagenet/calib_100")
else:
files = os.listdir(source_folder)
for filename in files:
if not filename.startswith("ILSVRC2012_val_") or not filename.endswith(".JPEG"):
continue
n_identifier = filename.split("_")[-1].split(".")[0]
folder_name = n_identifier
folder_path = os.path.join(source_folder, folder_name)
if not os.path.exists(folder_path):
os.makedirs(folder_path)
file_path = os.path.join(source_folder, filename)
destination = os.path.join(folder_path, filename)
shutil.move(file_path, destination)
print("File organization complete.")
if not os.path.exists(calib_data_path):
os.makedirs(calib_data_path)
subfolders = os.listdir(source_folder)
for subfolder in subfolders:
source_subfolder = os.path.join(source_folder, subfolder)
destination_subfolder = os.path.join(calib_data_path, subfolder)
os.makedirs(destination_subfolder, exist_ok=True)
files = os.listdir(source_subfolder)
if files:
file_to_copy = files[0]
source_file = os.path.join(source_subfolder, file_to_copy)
destination_file = os.path.join(destination_subfolder, file_to_copy)
shutil.copy(source_file, destination_file)
print("Creating calibration dataset complete.")
if not os.path.exists(source_folder):
print("The provided data path does not exist.")
sys.exit(1)
The storage format of the val_data of the ImageNet dataset organized as follows:
val_data
n01440764
ILSVRC2012_val_00000293.JPEG
ILSVRC2012_val_00002138.JPEG
…
n01443537
ILSVRC2012_val_00000236.JPEG
ILSVRC2012_val_00000262.JPEG
…
…
The storage format of the calib_data of the ImageNet dataset organized as follows:
calib_data
n01440764
ILSVRC2012_val_00000293.JPEG
n01443537
ILSVRC2012_val_00000236.JPEG
…
4) Quantization Procedure#
First, create a data reader that gathers calibration statistics from the target dataset. Next, inside quantize_model, constructing the quantized model and passing in your configuration.
# import copy
# import cv2
# import numpy as np
# import onnx
# import onnxruntime
# import torch
# import torchvision
# from onnxruntime.quantization import CalibrationDataReader
# from torchvision import transforms
# from quark.onnx import (
# AdaQuantConfig,
# AdaRoundConfig,
# BFloat16Spec,
# BFP16Spec,
# CalibMethod,
# CLEConfig,
# Int8Spec,
# Int16Spec,
# ModelQuantizer,
# QConfig,
# QLayerConfig,
# XInt8Spec,
# )
# class ImageDataReader(CalibrationDataReader):
# def __init__(self, calibration_image_folder: str, input_name: str):
# self.enum_data = None
# self.input_name = input_name
# self.data_list = self._preprocess_images(calibration_image_folder)
# def _preprocess_images(self, image_folder: str):
# data_list = []
# img_names = [f for f in os.listdir(image_folder) if f.endswith(".png") or f.endswith(".jpg") or f.endswith(".JPEG")]
# for name in img_names:
# input_image = cv2.imread(os.path.join(image_folder, name))
# # Resize the input image. Because the size of Resnet50 is 224.
# input_image = cv2.resize(input_image, (224, 224))
# input_data = np.array(input_image).astype(np.float32)
# # Customer Pre-Process
# input_data = input_data.transpose(2, 0, 1)
# input_size = input_data.shape
# if input_size[1] > input_size[2]:
# input_data = input_data.transpose(0, 2, 1)
# input_data = np.expand_dims(input_data, axis=0)
# input_data = input_data / 255.0
# data_list.append(input_data)
# return data_list
# def get_next(self):
# if self.enum_data is None:
# self.enum_data = iter([{self.input_name: data} for data in self.data_list])
# return next(self.enum_data, None)
# def rewind(self):
# self.enum_data = None
# from onnxruntime.quantization import CalibrationDataReader
import copy
import numpy as np
import onnxruntime
import torch
import torchvision
from onnxruntime.quantization.calibrate import CalibrationDataReader
from PIL import Image
from torchvision import transforms
from quark.onnx import (
AdaQuantConfig,
AdaRoundConfig,
BFloat16Spec,
BFP16Spec,
CalibMethod,
CLEConfig,
Int8Spec,
Int16Spec,
ModelQuantizer,
QConfig,
QLayerConfig,
XInt8Spec,
)
from quark.onnx.operators.custom_ops import get_library_path
# def get_model_input_name(input_model_path: str) -> str:
# model = onnx.load(input_model_path)
# model_input_name = model.graph.input[0].name
# return model_input_name
def _preprocess_images(images_folder: str, height: int, width: int, size_limit=0, batch_size=100):
"""
Loads a batch of images and preprocess them
parameter images_folder: path to folder storing images
parameter height: image height in pixels
parameter width: image width in pixels
parameter size_limit: number of images to load. Default is 0 which means all images are picked.
return: list of matrices characterizing multiple images
"""
image_path = os.listdir(images_folder)
image_names = []
for image_dir in image_path:
image_name = os.listdir(os.path.join(images_folder, image_dir))
image_names.append(os.path.join(image_dir, image_name[0]))
if size_limit > 0 and len(image_names) >= size_limit:
batch_filenames = [image_names[i] for i in range(size_limit)]
else:
batch_filenames = image_names
unconcatenated_batch_data = []
batch_data = []
for index, image_name in enumerate(batch_filenames):
image_filepath = images_folder + "/" + image_name
pillow_img = Image.new("RGB", (width, height))
pillow_img.paste(Image.open(image_filepath).resize((width, height)))
image_array = np.array(pillow_img) / 255.0
mean = np.array([0.485, 0.456, 0.406])
image_array = image_array - mean
std = np.array([0.229, 0.224, 0.225])
nchw_data = image_array / std
nchw_data = nchw_data.transpose((2, 0, 1))
nchw_data = np.expand_dims(nchw_data, axis=0)
nchw_data = nchw_data.astype(np.float32)
unconcatenated_batch_data.append(nchw_data)
if (index + 1) % batch_size == 0:
one_batch_data = np.concatenate(unconcatenated_batch_data, axis=0)
unconcatenated_batch_data.clear()
batch_data.append(one_batch_data)
return batch_data
class ImageDataReader(CalibrationDataReader):
def __init__(self, calibration_image_folder: str, model_path: str, data_size: int, batch_size: int):
self.enum_data = None
# Use inference session to get input shape.
session = onnxruntime.InferenceSession(model_path, providers=["CPUExecutionProvider"])
(_, _, height, width) = session.get_inputs()[0].shape
# Convert image to input data
self.nhwc_data_list = _preprocess_images(calibration_image_folder, height, width, data_size, batch_size)
self.input_name = session.get_inputs()[0].name
self.datasize = len(self.nhwc_data_list)
def get_next(self):
if self.enum_data is None:
self.enum_data = iter([{self.input_name: nhwc_data} for nhwc_data in self.nhwc_data_list])
return next(self.enum_data, None)
def rewind(self):
self.enum_data = None
def reset(self):
self.enum_data = None
As we plan to experiment with various methods and combinations, it is more efficient to organize them into groups and reference them by their labels when needed.
activation_spec_dict = {
"XINT8": XInt8Spec(),
"A8W8": Int8Spec(),
"A16W8": Int16Spec(),
"BF16": BFloat16Spec(),
"BFP16": BFP16Spec(),
}
weight_spec_dict = {
"XINT8": XInt8Spec(),
"A8W8": Int8Spec(),
"A16W8": Int8Spec(),
"BF16": BFloat16Spec(),
"BFP16": BFP16Spec(),
}
DEFAULT_ADAROUND_PARAMS = {
"DataSize": 1000,
"FixedSeed": 1705472343,
"BatchSize": 2,
"NumIterations": 1000,
"LearningRate": 0.1,
"OptimAlgorithm": "adaround",
"OptimDevice": "cpu",
"InferDevice": "cpu",
"EarlyStop": True,
}
DEFAULT_ADAQUANT_PARAMS = {
"DataSize": 1000,
"FixedSeed": 1705472343,
"BatchSize": 2,
"NumIterations": 1000,
"LearningRate": 0.00001,
"OptimAlgorithm": "adaquant",
"OptimDevice": "cpu",
"InferDevice": "cpu",
"EarlyStop": True,
}
Define a function to perform quantization based on your config
def quantize_model(args):
activation_spec = activation_spec_dict[args["config"]]
weight_spec = weight_spec_dict[args["config"]]
algo_confs = [CLEConfig()]
exclude_info = []
if args.get("exclude_nodes"):
exclude_nodes = args["exclude_nodes"].split(";")
exclude_nodes = [node_name.strip() for node_name in exclude_nodes]
exclude_info += exclude_nodes
if args["config"] == "XINT8":
enable_npucnn = True
else:
enable_npucnn = False
activation_spec.set_calibration_method(CalibMethod.MinMax)
weight_spec.set_calibration_method(CalibMethod.MinMax)
if args.get("learning_rate"):
lr = args["learning_rate"]
else:
if args.get("adround"):
lr = DEFAULT_ADAROUND_PARAMS["LearningRate"]
elif args.get("adaquant"):
lr = DEFAULT_ADAQUANT_PARAMS["LearningRate"]
if args.get("num_iters"):
num_iter = args["num_iters"]
else:
if args.get("adround"):
num_iter = DEFAULT_ADAROUND_PARAMS["NumIterations"]
elif args.get("adaquant"):
num_iter = DEFAULT_ADAQUANT_PARAMS["NumIterations"]
if args.get("adaround"):
adaround_algo = AdaRoundConfig(
data_size=DEFAULT_ADAROUND_PARAMS["DataSize"],
batch_size=DEFAULT_ADAROUND_PARAMS["BatchSize"],
num_iterations=num_iter,
learning_rate=lr,
early_stop=DEFAULT_ADAROUND_PARAMS["EarlyStop"],
)
algo_confs.append(adaround_algo)
if args.get("adaquant"):
adaround_algo = AdaQuantConfig(
data_size=DEFAULT_ADAQUANT_PARAMS["DataSize"],
batch_size=DEFAULT_ADAQUANT_PARAMS["BatchSize"],
num_iterations=num_iter,
learning_rate=lr,
early_stop=DEFAULT_ADAQUANT_PARAMS["EarlyStop"],
)
algo_confs.append(adaround_algo)
extra_info = {}
if args["config"] == "A8W8":
extra_info = {"EnableNPUCnn": enable_npucnn, "AlignSlice": False, "FoldRelu": True, "AlignConcat": True}
if args["config"] == "A16W8":
extra_info = {
"EnableNPUCnn": enable_npucnn,
"AlignSlice": False,
"FoldRelu": True,
"AlignConcat": True,
"AlignEltwiseQuantType": True,
}
if args["config"] == "BF16":
extra_info = {"EnableNPUCnn": enable_npucnn, "QuantizeAllOpTypes": True, "ForceQuantizeNoInputCheck": True}
quant_config = QConfig(
global_config=QLayerConfig(activation=activation_spec, weight=weight_spec),
algo_config=algo_confs,
use_external_data_format=args.get("save_as_external_data"),
exclude=exclude_info,
**extra_info,
)
# model_input_name = get_model_input_name(args["input_model_path"])
# calib_datareader = ImageDataReader(args["calib_data_path"], model_input_name)
calib_datareader = ImageDataReader(
args["calib_data_path"], args["input_model_path"], args["num_calib_data"], args["batch_size"]
)
quantizer = ModelQuantizer(quant_config)
quantizer.quantize_model(args["input_model_path"], args["output_model_path"], calib_datareader)
Create a basic configuration, then modify it based on each requirement
quant_config = {
"input_model_path": "models/resnet50-v1-12.onnx",
"calib_data_path": calib_data_path,
"num_calib_data": 1000,
"batch_size": 1,
"num_iters": 1,
"device": "cpu",
}
Run XINT8
quant_config_with_xint8 = copy.deepcopy(quant_config)
quant_config_with_xint8["output_model_path"] = "models/resnet152_xint8_quantized.onnx"
quant_config_with_xint8["config"] = "XINT8"
quantize_model(quant_config_with_xint8)
Run A8W8
quant_config_with_a8w8 = copy.deepcopy(quant_config)
quant_config_with_a8w8["output_model_path"] = "models/resnet152_a8w8_quantized.onnx"
quant_config_with_a8w8["config"] = "A8W8"
quantize_model(quant_config_with_a8w8)
Run A16W8
quant_config_with_a16w8 = copy.deepcopy(quant_config)
quant_config_with_a16w8["output_model_path"] = "models/resnet152_a16w8_quantized.onnx"
quant_config_with_a16w8["config"] = "A16W8"
quantize_model(quant_config_with_a16w8)
Run BF16
quant_config_with_bf16 = copy.deepcopy(quant_config)
quant_config_with_bf16["output_model_path"] = "models/resnet152_bf16_quantized.onnx"
quant_config_with_bf16["config"] = "BF16"
quantize_model(quant_config_with_bf16)
Run XINT8 with CLE
quant_config_with_xint8_cle = copy.deepcopy(quant_config)
quant_config_with_xint8_cle["output_model_path"] = "models/resnet152_xint8_cle_quantized.onnx"
quant_config_with_xint8_cle["config"] = "XINT8"
quant_config_with_xint8_cle["cle"] = True
quantize_model(quant_config_with_xint8_cle)
Run XINT8 with AdaRound
quant_config_with_xint8_adaround = copy.deepcopy(quant_config)
quant_config_with_xint8_adaround["output_model_path"] = "models/resnet152_xint8_adaround_quantized.onnx"
quant_config_with_xint8_adaround["config"] = "XINT8"
quant_config_with_xint8_adaround["adaround"] = True
quant_config_with_xint8_adaround["learning_rate"] = 0.1
quant_config_with_xint8_adaround["num_iters"] = 3000
quantize_model(quant_config_with_xint8_adaround)
Run XINT8 with AdaQuant
quant_config_with_xint8_adaquant = copy.deepcopy(quant_config)
quant_config_with_xint8_adaquant["output_model_path"] = "models/resnet152_xint8_adaquant_quantized.onnx"
quant_config_with_xint8_adaquant["config"] = "XINT8"
quant_config_with_xint8_adaquant["adaquant"] = True
quant_config_with_xint8_adaquant["learning_rate"] = 0.00001
quant_config_with_xint8_adaquant["num_iters"] = 10000
quantize_model(quant_config_with_xint8_adaquant)
Run XINT8 with Exclude Nodes
quant_config_with_xint8_exclude_nodes = copy.deepcopy(quant_config)
quant_config_with_xint8_exclude_nodes["output_model_path"] = "models/resnet152_xint8_exclude_nodes_quantized.onnx"
quant_config_with_xint8_exclude_nodes["config"] = "XINT8"
quant_config_with_xint8_exclude_nodes["exclude_nodes"] = "resnetv17_conv0_fwd; resnetv17_stage1_conv0_fwd"
quantize_model(quant_config_with_xint8_exclude_nodes)
5) Evaluation and Expected Results#
Evaluation is performed on the ImageNet validation set. We compare three models — (1) full-precision, (2) quantized with minmax calibration, and (3) quantized with layerwise calibration — to assess layerwise calibration’s effectiveness. The full-precision model serves as the baseline for measuring any accuracy change caused by quantization.
ImageNet has 1,000 classes, so we report both Prec@1 and Prec@5 to capture strict and relaxed accuracy. Both metrics are reported as percentages (higher is better). Prec@1 shows exact single-label correctness; Prec@5 is useful on large, fine-grained label spaces because it captures near-misses where the correct class is among the model’s top candidates.
class AverageMeter:
"""Computes and stores the average and current value"""
def __init__(self):
self.reset()
def reset(self):
self.val = 0
self.avg = 0
self.sum = 0
self.count = 0
def update(self, val, n=1):
self.val = val
self.sum += val * n
self.count += n
self.avg = self.sum / self.count
def load_loader(data_dir, batch_size, workers):
data_transform = transforms.Compose(
[
transforms.Resize(256),
transforms.CenterCrop(224),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
]
)
dataset = torchvision.datasets.ImageFolder(data_dir, data_transform)
data_loader = torch.utils.data.DataLoader(
dataset, batch_size=batch_size, shuffle=False, num_workers=workers, pin_memory=True
)
return data_loader
def accuracy_np(output, target):
max_indices = np.argsort(output, axis=1)[:, ::-1]
top5 = 100 * np.equal(max_indices[:, :5], target[:, np.newaxis]).sum(axis=1).mean()
top1 = 100 * np.equal(max_indices[:, 0], target).mean()
return top1, top5
def metrics(onnx_model_path, sess_options, providers, data_loader, print_freq):
session = onnxruntime.InferenceSession(onnx_model_path, sess_options, providers=providers)
input_name = session.get_inputs()[0].name
batch_time = AverageMeter()
top1 = AverageMeter()
top5 = AverageMeter()
end = time.time()
for i, (input, target) in enumerate(data_loader):
# run the net and return prediction
output = session.run([], {input_name: input.data.numpy()})
output = output[0]
# measure accuracy and record loss
prec1, prec5 = accuracy_np(output, target.numpy())
top1.update(prec1.item(), input.size(0))
top5.update(prec5.item(), input.size(0))
# measure elapsed time
batch_time.update(time.time() - end)
end = time.time()
if i % print_freq == 0:
print(
f"Test: [{i}/{len(data_loader)}]\t"
f"Time {batch_time.val:.3f} ({batch_time.avg:.3f}, {input.size(0) / batch_time.avg:.3f}/s, "
f"{100 * batch_time.avg / input.size(0):.3f} ms/sample) \t"
f"Prec@1 {top1.val:.3f} ({top1.avg:.3f})\t"
f"Prec@5 {top5.val:.3f} ({top5.avg:.3f})"
)
return top1, top5
def evaluate(args: dict):
args["gpu_id"] = 0
# Set graph optimization level
sess_options = onnxruntime.SessionOptions()
sess_options.graph_optimization_level = onnxruntime.GraphOptimizationLevel.ORT_ENABLE_ALL
if args.get("profile"):
sess_options.enable_profiling = True
if args.get("onnx_output_opt"):
sess_options.optimized_model_filepath = args["onnx_output_opt"]
if args.get("gpu"):
if "ROCMExecutionProvider" in onnxruntime.get_available_providers():
device = "ROCM"
providers = ["ROCMExecutionProvider"]
elif "CUDAExecutionProvider" in onnxruntime.get_available_providers():
device = "CUDA"
providers = ["CUDAExecutionProvider"]
else:
device = "CPU"
providers = ["CPUExecutionProvider"]
print("Warning: GPU is not available, use CPU instead.")
else:
device = "CPU"
providers = ["CPUExecutionProvider"]
sess_options.register_custom_ops_library(get_library_path(device))
if args.get("onnx_input"):
val_loader = load_loader(args["data"], args["batch_size"], args["workers"])
f_top1, f_top5 = metrics(args["onnx_input"], sess_options, providers, val_loader, args["print_freq"])
print(f" * Prec@1 {f_top1.avg:.3f} ({100 - f_top1.avg:.3f}) Prec@5 {f_top5.avg:.3f} ({100.0 - f_top5.avg:.3f})")
elif args.get("onnx_float") and args.get("onnx_quant"):
val_loader = load_loader(args["data"], args["batch_size"], args["workers"])
f_top1, f_top5 = metrics(args["onnx_float"], sess_options, providers, val_loader, args["print_freq"])
f_top1 = format(f_top1.avg, ".2f")
f_top5 = format(f_top5.avg, ".2f")
q_top1, q_top5 = metrics(args["onnx_quant"], sess_options, providers, val_loader, args["print_freq"])
q_top1 = format(q_top1.avg, ".2f")
q_top5 = format(q_top5.avg, ".2f")
f_size = format(os.path.getsize(args["onnx_float"]) / (1024 * 1024), ".2f")
q_size = format(os.path.getsize(args["onnx_quant"]) / (1024 * 1024), ".2f")
"""
--------------------------------------------------------
| | float model | quantized model |
--------------------------------------------------------
| **** | **** | **** |
--------------------------------------------------------
| Model Size | **** | **** |
--------------------------------------------------------
"""
from rich.console import Console
from rich.table import Table
console = Console()
table = Table()
table.add_column("")
table.add_column("Float Model")
table.add_column("Quantized Model", style="bold green1")
table.add_row("Model", args["onnx_float"], args["onnx_quant"])
table.add_row("Model Size", str(f_size) + " MB", str(q_size) + " MB")
table.add_row("Prec@1", str(f_top1) + " %", str(q_top1) + " %")
table.add_row("Prec@5", str(f_top5) + " %", str(q_top5) + " %")
console.print(table)
else:
print("Please specify both model-float and model-quant or model-input for evaluation.")
First, define an evaluation config, and record accuracy of the Full Precision model on ImageNet val dataset
eval_config = {
"data": source_folder,
"batch_size": 100,
"workers": 1,
"gpu": False,
"print_freq": 1000,
}
full_precision_eval_config = copy.deepcopy(eval_config)
full_precision_eval_config["onnx_input"] = "models/resnet50-v1-12.onnx"
evaluate(full_precision_eval_config)
We have quantized several models, but to keep this tutorial simple, we will focus on evaluating the A8W8 model as an example. If you are interested in exploring other quantization configurations, you can follow the provided code and run your own experiments. For now, let’s specify the path to the A8W8 quantized model and measure its accuracy on the ImageNet validation dataset.
quant_eval_config = copy.deepcopy(eval_config)
quant_eval_config["onnx_input"] = "models/resnet152_a8w8_quantized.onnx"
evaluate(quant_eval_config)
The following table contains the expected results, but please note that different machines can lead to minor variations in the accuracy of quantized model.
Float Model |
Quantized Model |
|
|---|---|---|
Model Size |
97.82 MB |
25.62 MB |
Prec@1 |
74.114 % |
73.562 % |
Prec@5 |
91.716 % |
91.420 % |