vai_q_tensorflow Supported Operations and APIs - 3.5 English

Vitis AI User Guide (UG1414)

Document ID
UG1414
Release Date
2023-09-28
Version
3.5 English

The following table lists the supported vai_q_tensorflow operations and APIs.

Table 1. Supported Operations and APIs for vai_q_tensorflow
Type Operation Type tf.nn tf.layers tf.keras.layers
Convolution Conv2D

DepthwiseConv2dNative

atrous_conv2d

conv2d

conv2d_transpose

depthwise_conv2d_native

separable_conv2d

Conv2D

Conv2DTranspose

SeparableConv2D

Conv2D

Conv2DTranspose

DepthwiseConv2D

SeparaleConv2D

Fully Connected MatMul / Dense Dense
BiasAdd BiasAdd

Add

bias_add / /
Pooling AvgPool

Mean

MaxPool

avg_pool

max_pool

AveragePooling2D

MaxPooling2D

AveragePooling2D

MaxPool2D

Activation ReLU

ReLU6

Sigmoid

Swish

Hard-sigmoid

Hard-swish

relu

relu6

leaky_relu

swish

/ ReLU

Leaky ReLU

BatchNorm[#1] FusedBatchNorm batch_normalization

batch_norm_with_global_normalization

fused_batch_norm

BatchNormalization BatchNormalization
Upsampling ResizeBilinear

ResizeNearestNeighbor

/ / UpSampling2D
Concat Concat

ConcatV2

/ / Concatenate
Others Placeholder

Const

Pad

Squeeze

Reshape

ExpandDims

Max

Transpose

dropout[#2]

softmax[#3]

depth_to_space

Dropout[#2]

Flatten

Input

Flatten

Reshape

Zeropadding2D

Softmax

  1. Only supports Conv2D/DepthwiseConv2D/Dense+BN. BN is folded to increase inference performance.
  2. Dropout is deleted to increase inference performance.
  3. vai_q_tensorflow does not quantize the softmax output.
Note: The list of operators supported by the Vitis AI quantizer is not the only limiting factor for model deployment, and users should also review operator support for their selected DPU architecture. Operators not supported by the DPU can be executed on the CPU. For more information, see Supported Operators and DPU Limitations.