CNN2SNN toolkit

Overview

The Brainchip CNN2SNN toolkit provides a means to convert a quantized model obtained using QuantizeML to a low-latency and low-power network for use with the Akida runtime.

Conversion flow

CNN2SNN offers a simple convert function that takes a quantized model as input and converts it into an Akida runtime compatible network.

Let’s take the DS-CNN model from our zoo that targets KWS task as an example:

from akida_models import ds_cnn_kws_pretrained
from cnn2snn import convert

# Load a pretrained 8/4/4 quantized model
quantized_model = ds_cnn_kws_pretrained()
model_akida = convert(quantized_model)

Conversion compatibility

It is possible to check if a float model is compatible with Akida conversion using the check_model_compatibility helper. This helper will check that the model quantization scheme is allowed and that building blocks are compatible with Akida layers blocks, convert the model and optionally map on an Akida hardware.

Command-line interface

In addition to the CNN2SNN programming API, the CNN2SNN toolkit provides a command-line interface to perform conversion to an Akida runtime compatible model. Converting a quantized model into an Akida model using the CLI makes use of the convert function.

Examples

Convert the DS-CNN/KWS 8/4/4 quantized model:

wget https://data.brainchip.com/models/AkidaV2/ds_cnn/ds_cnn_kws_i8_w4_a4.h5

cnn2snn convert -m ds_cnn_kws_i8_w4_a4.h5

An Akida .fbz model named ds_cnn_kws_i8_w4_a4.fbz is then saved. This model can be loaded back into an akida.Model and run on Akida runtime.

Deprecated CLI actions

The scale and shift options of the convert CLI action that were used to set input scaling parameters are now deprecated.

CNN2SNN CLI comes with additional actions that are also deprecated and should no longer be used: quantize, reshape and calibrate. You should rather use QuantizeML tool to perform the same operations.

Handling Akida 1.0 and Akida 2.0 specificities

Conversion towards Akida 1.0 or Akida 2.0 is inherently different because the targeted SoC or IP comes with different features. By default, a model is converted towards Akida 2.0. It is however possible to convert towards Akida 1.0.

Using the programming interface:

from akida_models import ds_cnn_kws_pretrained
from cnn2snn import convert, set_akida_version, AkidaVersion

with set_akida_version(AkidaVersion.v1):
    quantized_model = ds_cnn_kws_pretrained()
    model_akida = convert(quantized_model)

Using the CLI interface:

wget https://data.brainchip.com/models/AkidaV1/ds_cnn/ds_cnn_kws_iq8_wq4_aq4_laq1.h5

CNN2SNN_TARGET_AKIDA_VERSION=v1 cnn2snn convert -m ds_cnn_kws_iq8_wq4_aq4_laq1.h5

Note

Legacy quantization API

Warning

While it is possible to quantize Akida 1.0 models using cnn2snn legacy quantization blocks, such usage is deprecated. You should rather use QuantizeML tool to quantize a model whenever possible.

Typical quantization scenario

The CNN2SNN toolkit offers a turnkey solution to quantize a model: the quantize function. It replaces the neural Keras layers (Conv2D, SeparableConv2D and Dense) and the ReLU layers with custom CNN2SNN layers, which are Quantization Aware derived versions of the base Keras layer types. The obtained quantized model is still a Keras model with a mix of CNN2SNN quantized layers (QuantizedReLU, QuantizedDense, etc.) and standard Keras layers (BatchNormalization, MaxPool2D, etc.).

Direct quantization of a standard Keras model (also called post-training quantization) generally introduces a drop in performance. This drop is usually small for 8-bit or even 4-bit quantization of simple models, but it can be very significant for low quantization bitwidth and complex models.

If the quantized model offers acceptable performance, it can be directly converted into an Akida model, ready to be loaded on the Akida NSoC (see the convert function).

However, if the performance drop is too high, a Quantization Aware Training is required to recover the performance prior to quantization. Since the quantized model is a Keras model, it can then be trained using the standard Keras API.

Note that quantizing directly to the target bitwidth is not mandatory: it is possible to proceed with quantization in a serie of smaller steps. For example, it may be beneficial to keep float weights and only quantize activations, retrain, and then, quantize weights.

Design compatibility constraints

When designing a tf.keras model, consider design compatibility at these distinct levels before the quantization stage:

  • Only serial and feedforward arrangements can be converted1.

  • Supported Keras layers are listed below.

  • Order of the layers is important, e.g. a BatchNormalization layer must be placed before the activation, and not after.

  • Some constraints are needed about layer’s parameters, e.g. a MaxPool2D layer must have the same padding as its corresponding convolutional layer.

All these design compatibility constraints are summarized in the 1.0 specific cnn2snn.compatibility_checks.check_model_compatibility function. A good practice is to check model compatibility before going through the training process 2.

Helpers (see Layer Blocks) are available in the akida_models PyPI package to easily create a compatible model from scratch.

Command-line interface

In addition to the cnn2snn programming API, the CNN2SNN toolkit also provides a command-line interface to perform quantization, conversion to an Akida NSoC compatible model or model reshape.

Quantizing a standard Keras model or a CNN2SNN quantized model using the CLI makes use of the cnn2snn.quantize Python function. The same arguments, i.e. the quantization bitwidths for weights and activations, are required.

Examples

Quantize a standard Keras model with 4-bit weights and activations and 8-bit input weights:

cnn2snn quantize -m model_keras.h5 -wq 4 -aq 4 -iq 8

The quantized model is automatically saved to model_keras_iq8_wq4_aq4.h5.

Quantize an already quantized model with different quantization bitwidths:

cnn2snn quantize -m model_keras_iq8_wq4_aq4.h5 -wq 2 -aq 2

A new quantized model named model_keras_iq2_wq2_aq2.h5 is saved.

A model can be reshaped (change of input shape) using CNN2SNN CLI that makes use of the cnn2snn.transforms.reshape function. This will only apply to Sequential models, a sequentialize helper is provided for convenience.

Examples

Reshape a model to 160x96:

cnn2snn reshape -m model_keras.h5 -iw 160 -ih 96

A reshaped model will be saved as model_keras_160_96.h5.

Layers Considerations

Supported layer types

The CNN2SNN toolkit provides quantization of Keras models with the following Keras layer types:

  • Core Neural Layers:

  • Specialized Layers:

  • Other Layers (from tf.keras):

    • ReLU

    • BatchNormalization

    • MaxPooling2D

    • GlobalAveragePooling2D

    • Dropout

    • Flatten

    • Reshape

    • Input

CNN2SNN Quantization Aware layers

Several articles have reported4 that the quantization of a pre-trained float Keras model using 8-bit precision can be performed with a minimal loss of accuracy for simple models, but that for lower bitwidth or complex models a Quantization Aware Training of the quantized model may be required.

The CNN2SNN toolkit therefore includes Quantization Aware versions of the base Keras layers.

These layers are produced when quantizing a standard Keras model using the quantize function: it replaces the base Keras layers with their Quantization Aware counterparts (see the quantize function).

Quantization Aware Training simulates the effect of quantization in the forward pass, yet using a straight-through estimator for the quantization gradient in the backward pass. For the stochastic gradient descent to be efficient, the weights are stored as float values and updated with high precision during back propagation. This ensures sufficient precision in accumulating tiny weights adjustments.

The CNN2SNN toolkit includes two classes of Quantization Aware layers:

Most of the parameters for the quantized processing layers are identical to those used when defining a model using standard Keras layers. However, each of these layers also includes a quantizer parameter that specifies the WeightQuantizer object to use during the Quantization Aware Training.

The quantized ReLU takes a single parameter corresponding to the bitwidth of the quantized activations.

Training-Only Layers

Training is done within the Keras environment and training-only layers may be added at will, such as BatchNormalization or Dropout layers. These are handled fully by Keras during the training and do not need to be modified to be Akida-compatible for inference.

As regards the implementation within the Akida neuromorphic IP: it may be helpful to understand that the associated scaling operations (multiplication and shift) are never performed during inference. The computational cost is reduced by wrapping the (optional) batch normalization function and quantized activation function into the spike generating thresholds and other parameters of the Akida model. That process is completely transparent to the user. It does, however, have an important consequence for the output of the final layer of the model; see Final Layers below.

First Layers

Most layers of an Akida model only accept sparse inputs. In order to support the most common classes of models in computer vision, a special layer (InputConvolutional) is however able to receive image data (8-bit grayscale or RGB). See the Akida user guide for further details.

The CNN2SNN toolkit supports any Quantization Aware Training layer as the first layer in the model. The type of input accepted by that layer can be specified during conversion, but only models starting with a QuantizedConv2D layer will accept dense inputs, thanks to the special InputConvolutional layer.

Input Scaling

The InputConvolutional layer only receives 8-bit input values:

  • if the data is already in 8-bit format it can be sent to the Akida inputs without rescaling.

  • if the data has been scaled to ease training, it is necessary to provide the scaling coefficients at model conversion.

This applies to the common case where input data are natively 8-bit. If input data are not 8-bit, the process is more complex, and we recommend applying rescaling in two steps:

  1. Taking the data to an 8-bit unsigned integer format suitable for Akida architecture. Apply this step both for training and inference data.

  2. Rescaling the 8-bit values to some unit or zero centered range suitable for CNN training, as above. This step should only be applied for the CNN training. Also, remember to provide those scaling coefficients when converting the trained model to an Akida-compatible format.

Final Layers

As is typical for CNNs, the final layer of a model does not include the standard activation nonlinearity. If that is the case, once converted to Akida hardware, the model will give the potentials levels and in most cases, taking the maximum among these values is sufficient to obtain the correct response from the model. However, if there is a difference in performance between the Keras and the Akida-compatible implementations of the model, it is likely be at this step.

Tips and Tricks

In some cases, it may be useful to adapt existing CNN models in order to simplify or enhance the converted model. Here’s a short list of some possible substitutions that might come in handy:


1

Parallel layers and “residual” connections are currently not supported.

2

Check model compatibility must be applied on a quantized model. It then requires to quantize the model first.

3

The spike value depends on the intensity of the potential, see the Akida documentation for details on the activation.

4

See for instance “Quantizing deep convolutional networks for efficient inference: A whitepaper” - Raghuraman Krishnamoorthi, 2018