Skip to content
This repository has been archived by the owner on Nov 17, 2023. It is now read-only.

Commit

Permalink
[MKLDNN]Enhance Quantization APIs and Tutorial (#15448)
Browse files Browse the repository at this point in the history 
* enhance api and new tutorial

* Update MKLDNN_QUANTIZATION.md

update

* fix lint

* modify pics

* skip test

* add quantize layer in graph

* update

* remove center css flag

* change requantize color

* fix markdown pics

* change to use png

* Update MKLDNN_QUANTIZATION.md

update

* enable ipython script

* fix png

* fix lint

* Update MKLDNN_QUANTIZATION.md

* change title

* trigger

* use lower case

* some typo

* some typo

* use dmlc web data

* trigger

* trigger
  • Loading branch information
@xinyu-intel @pengzhao-intel
xinyu-intel authored and pengzhao-intel committed Aug 1, 2019
1 parent 773f4dc commit b3064c5
Show file tree
Hide file tree
Showing 6 changed files with 657 additions and 108 deletions.
3 changes: 2 additions & 1 deletion docs/tutorials/index.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -105,6 +105,8 @@ Select API: 
* [Module to Gluon API](/tutorials/python/module_to_gluon.html)
* [Gluon end to end from training to inference](/tutorials/gluon/gluon_from_experiment_to_deployment.html)
* [Automatic Mixed Precision in Gluon](/tutorials/amp/amp_tutorial.html)
* [How to build and install MXNet with MKL-DNN backend](/tutorials/mkldnn/MKLDNN_README.html)
* [How to quantize custom models with MKL-DNN backend](/tutorials/mkldnn/mkldnn_quantization.html)<span style="color:red"> (new!) </span>
* API Guides
* Core APIs
* NDArray
Expand Down Expand Up @@ -157,7 +159,6 @@ Select API:&nbsp;
* [Large-Scale Multi-Host Multi-GPU Image Classification](/tutorials/vision/large_scale_classification.html)
* [Importing an ONNX model into MXNet](/tutorials/onnx/super_resolution.html)
* [Optimizing Deep Learning Computation Graphs with TensorRT](/tutorials/tensorrt/inference_with_trt.html)
* [How to build and install MXNet with MKL-DNN backend](/tutorials/mkldnn/MKLDNN_README.html)
* API Guides
* Core APIs
* NDArray
Expand Down
259 changes: 259 additions & 0 deletions docs/tutorials/mkldnn/mkldnn_quantization.md
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,259 @@

<!--- Licensed to the Apache Software Foundation (ASF) under one -->
<!--- or more contributor license agreements. See the NOTICE file -->
<!--- distributed with this work for additional information -->
<!--- regarding copyright ownership. The ASF licenses this file -->
<!--- to you under the Apache License, Version 2.0 (the -->
<!--- "License"); you may not use this file except in compliance -->
<!--- with the License. You may obtain a copy of the License at -->

<!--- http://www.apache.org/licenses/LICENSE-2.0 -->

<!--- Unless required by applicable law or agreed to in writing, -->
<!--- software distributed under the License is distributed on an -->
<!--- "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY -->
<!--- KIND, either express or implied. See the License for the -->
<!--- specific language governing permissions and limitations -->
<!--- under the License. -->

# Quantize custom models with MKL-DNN backend

This document is to introduce how to quantize the customer models from FP32 to INT8 with Apache/MXNet toolkit and APIs under Intel CPU.

If you are not familiar with Apache/MXNet quantization flow, please reference [quantization blog](https://medium.com/apache-mxnet/model-quantization-for-production-level-neural-network-inference-f54462ebba05) first, and the performance data is shown in [Apache/MXNet C++ interface](https://github.com/apache/incubator-mxnet/tree/master/cpp-package/example/inference) and [GluonCV](https://gluon-cv.mxnet.io/build/examples_deployment/int8_inference.html).

## Installation and Prerequisites

Installing MXNet with MKLDNN backend is an easy and essential process. You can follow [How to build and install MXNet with MKL-DNN backend](https://mxnet.incubator.apache.org/tutorials/mkldnn/MKLDNN_README.html) to build and install MXNet from source. Also, you can install the release or nightly version via PyPi and pip directly by running:

```
# release version
pip install mxnet-mkl
# nightly version
pip install mxnet-mkl --pre
```

## Image Classification Demo

A quantization script [imagenet_gen_qsym_mkldnn.py](https://github.com/apache/incubator-mxnet/blob/master/example/quantization/imagenet_gen_qsym_mkldnn.py) has been designed to launch quantization for image-classification models. This script is integrated with [Gluon-CV modelzoo](https://gluon-cv.mxnet.io/model_zoo/classification.html), so that all pre-trained models can be downloaded from Gluon-CV and then converted for quantization. For details, you can refer [Model Quantization with Calibration Examples](https://github.com/apache/incubator-mxnet/blob/master/example/quantization/README.md).

## Integrate Quantization Flow to Your Project

Quantization flow works for both symbolic and Gluon models. If you're using Gluon, you can first refer [Saving and Loading Gluon Models](https://mxnet.incubator.apache.org/versions/master/tutorials/gluon/save_load_params.html) to hybridize your computation graph and export it as a symbol before running quantization.

In general, the quantization flow includes 4 steps. The user can get the acceptable accuracy from step 1 to 3 with minimum effort. Most of thing in this stage is out-of-box and the data scientists and researchers only need to focus on how to represent data and layers in their model. After a quantized model is generated, you may want to deploy it online and the performance will be the next key point. Thus, step 4, calibration, can improve the performance a lot by reducing lots of runtime calculation.

![quantization flow](https://github.com/dmlc/web-data/raw/master/mxnet/tutorials/mkldnn/quantization/quantization.png)

Now, we are going to take Gluon ResNet18 as an example to show how each step work.

### Initialize Model

```python
import logging
import mxnet as mx
from mxnet.gluon.model_zoo import vision
from mxnet.contrib.quantization import *

logging.basicConfig()
logger = logging.getLogger('logger')
logger.setLevel(logging.INFO)

batch_shape = (1, 3, 224, 224)
resnet18 = vision.resnet18_v1(pretrained=True)
resnet18.hybridize()
resnet18.forward(mx.nd.zeros(batch_shape))
resnet18.export('resnet18_v1')
sym, arg_params, aux_params = mx.model.load_checkpoint('resnet18_v1', 0)
# (optional) visualize float32 model
mx.viz.plot_network(sym)
```
First, we download resnet18-v1 model from gluon modelzoo and export it as a symbol. You can visualize float32 model. Below is a raw residual block.

![float32 model](https://github.com/dmlc/web-data/raw/master/mxnet/tutorials/mkldnn/quantization/fp32_raw.png)

#### Model Fusion

```python
sym = sym.get_backend_symbol('MKLDNN_QUANTIZE')
# (optional) visualize fused float32 model
mx.viz.plot_network(sym)
```
It's important to add this line to enable graph fusion before quantization to get better performance. Below is a fused residual block. Batchnorm, Activation and elemwise_add are fused into Convolution.

![float32 fused model](https://github.com/dmlc/web-data/raw/master/mxnet/tutorials/mkldnn/quantization/fp32_fusion.png)

### Quantize Model

A python interface `quantize_graph` is provided for the user. Thus, it is very flexible for the data scientist to construct the expected models based on different requirements in a real deployment.

```python
# quantize configs
# set exclude layers
excluded_names = []
# set calib mode.
calib_mode = 'none'
# set calib_layer
calib_layer = None
# set quantized_dtype
quantized_dtype = 'auto'
logger.info('Quantizing FP32 model Resnet18-V1')
qsym, qarg_params, aux_params, collector = quantize_graph(sym=sym, arg_params=arg_params, aux_params=aux_params,
excluded_sym_names=excluded_names,
calib_mode=calib_mode, calib_layer=calib_layer,
quantized_dtype=quantized_dtype, logger=logger)
# (optional) visualize quantized model
mx.viz.plot_network(qsym)
# save quantized model
mx.model.save_checkpoint('quantized-resnet18_v1', 0, qsym, qarg_params, aux_params)
```

By applying `quantize_graph` to the symbolic model, a new quantized model can be generated, named `qsym` along with its parameters. We can see `_contrib_requantize` operators are inserted after `Convolution` to convert the INT32 output to FP32.

![none calibrated model](https://github.com/dmlc/web-data/raw/master/mxnet/tutorials/mkldnn/quantization/none_calib.png)

Below table gives some descriptions.

| param | type | description|
|--------------------|-----------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| excluded_sym_names | list of strings | A list of strings representing the names of the symbols that users want to excluding from being quantized.|
| calib_mode | str | If calib_mode='none', no calibration will be used and the thresholds for requantization after the corresponding layers will be calculated at runtime by calling min and max operators. The quantized models generated in this mode are normally 10-20% slower than those with calibrations during inference.<br>If calib_mode='naive', the min and max values of the layer outputs from a calibration dataset will be directly taken as the thresholds for quantization.<br>If calib_mode='entropy', the thresholds for quantization will be derived such that the KL divergence between the distributions of FP32 layer outputs and quantized layer outputs is minimized based upon the calibration dataset. |
| calib_layer | function | Given a layer's output name in string, return True or False for deciding whether to calibrate this layer.<br>If yes, the statistics of the layer's output will be collected; otherwise, no information of the layer's output will be collected.<br>If not provided, all the layers' outputs that need requantization will be collected.|
| quantized_dtype | str | The quantized destination type for input data. Currently support 'int8', 'uint8' and 'auto'.<br>'auto' means automatically select output type according to calibration result.|

### Evaluate & Tune

Now, you get a pair of quantized symbol and params file for inference. For Gluon inference, only difference is to load model and params by a SymbolBlock as below example:

```python
quantized_net = mx.gluon.SymbolBlock.imports('quantized-resnet18_v1-symbol.json', 'data', 'quantized-resnet18_v1-0000.params')
quantized_net.hybridize(static_shape=True, static_alloc=True)
batch_size = 1
data = mx.nd.ones((batch_size,3,224,224))
quantized_net(data)
```

Now, you can get the accuracy from a quantized network. Furthermore, you can try to select different layers or OPs to be quantized by `excluded_sym_names` parameter and figure out an acceptable accuracy.

### Calibrate Model (optional for performance)

The quantized model generated in previous steps can be very slow during inference since it will calculate min and max at runtime. We recommend using offline calibration for better performance by setting `calib_mode` to `naive` or `entropy`. And then calling `set_monitor_callback` api to collect layer information with a subset of the validation datasets before int8 inference.

```python
# quantization configs
# set exclude layers
excluded_names = []
# set calib mode.
calib_mode = 'naive'
# set calib_layer
calib_layer = None
# set quantized_dtype
quantized_dtype = 'auto'
logger.info('Quantizing FP32 model resnet18-V1')
cqsym, cqarg_params, aux_params, collector = quantize_graph(sym=sym, arg_params=arg_params, aux_params=aux_params,
excluded_sym_names=excluded_names,
calib_mode=calib_mode, calib_layer=calib_layer,
quantized_dtype=quantized_dtype, logger=logger)

# download imagenet validation dataset
mx.test_utils.download('http://data.mxnet.io/data/val_256_q90.rec', 'dataset.rec')
# set rgb info for data
mean_std = {'mean_r': 123.68, 'mean_g': 116.779, 'mean_b': 103.939, 'std_r': 58.393, 'std_g': 57.12, 'std_b': 57.375}
# set batch size
batch_size = 16
# create DataIter
data = mx.io.ImageRecordIter(path_imgrec='dataset.rec', batch_size=batch_size, data_shape=batch_shape[1:], rand_crop=False, rand_mirror=False, **mean_std)
# create module
mod = mx.mod.Module(symbol=sym, label_names=None, context=mx.cpu())
mod.bind(for_training=False, data_shapes=data.provide_data, label_shapes=None)
mod.set_params(arg_params, aux_params)

# calibration configs
# set num_calib_batches
num_calib_batches = 5
max_num_examples = num_calib_batches * batch_size
# monitor FP32 Inference
mod._exec_group.execs[0].set_monitor_callback(collector.collect, monitor_all=True)
num_batches = 0
num_examples = 0
for batch in data:
mod.forward(data_batch=batch, is_train=False)
num_batches += 1
num_examples += batch_size
if num_examples >= max_num_examples:
break
if logger is not None:
logger.info("Collected statistics from %d batches with batch_size=%d"
% (num_batches, batch_size))
```

After that, layer information will be filled into the `collector` returned by `quantize_graph` api. Then, you need to write the layer information into int8 model by calling `calib_graph` api.


```python
# write scaling factor into quantized symbol
cqsym, cqarg_params, aux_params = calib_graph(qsym=cqsym, arg_params=arg_params, aux_params=aux_params,
collector=collector, calib_mode=calib_mode,
quantized_dtype=quantized_dtype, logger=logger)
# (optional) visualize quantized model
mx.viz.plot_network(cqsym)
```

Below is a quantized residual block with naive calibration. We can see `min_calib_range` and `max_calib_range` are written into `_contrib_requantize` operators.

![naive calibrated model](https://github.com/dmlc/web-data/raw/master/mxnet/tutorials/mkldnn/quantization/naive_calib.png)

When you get a quantized model with calibration, keeping sure to call fusion api again since this can fuse some `requantize` or `dequantize` operators for further performance improvement.

```python
# perform post-quantization fusion
cqsym = cqsym.get_backend_symbol('MKLDNN_QUANTIZE')
# (optional) visualize post-quantized model
mx.viz.plot_network(cqsym)
# save quantized model
mx.model.save_checkpoint('quantized-resnet18_v1', 0, cqsym, cqarg_params, aux_params)
```

Below is a post-quantized residual block. We can see `_contrib_requantize` operators are fused into `Convolution` operators.

![post-quantized model](https://github.com/dmlc/web-data/raw/master/mxnet/tutorials/mkldnn/quantization/post_quantize.png)

BTW, You can also modify the `min_calib_range` and `max_calib_range` in the JSON file directly.

```
{
"op": "_sg_mkldnn_conv",
"name": "quantized_sg_mkldnn_conv_bn_act_6",
"attrs": {
"max_calib_range": "3.562147",
"min_calib_range": "0.000000",
"quantized": "true",
"with_act": "true",
"with_bn": "true"
},
......
```

### Tips for Model Calibration

#### Accuracy Tuning

- Try to use `entropy` calib mode;

- Try to exclude some layers which may cause obvious accuracy drop;

- Change calibration dataset by setting different `num_calib_batches` or shuffle your validation dataset;

#### Performance Tuning

- Keep sure to perform graph fusion before quantization;

- If lots of `requantize` layers exist, keep sure to perform post-quantization fusion after calibration;

- Compare the MXNet profile or `MKLDNN_VERBOSE` of float32 and int8 inference;

## Deploy with Python/C++

MXNet also supports deploy quantized models with C++. Refer [MXNet C++ Package](https://github.com/apache/incubator-mxnet/blob/master/cpp-package/README.md) for more details.

<!-- INSERT SOURCE DOWNLOAD BUTTONS -->
Loading

0 comments on commit b3064c5

Please sign in to comment.