Operator lists¶
MACE Micro supported operators¶
Float32 and bfloat16 operators
- batch_norm
- conv_2d
- depthwise_conv_2d
- pooling
- activation
- argmax
- bias_add
- cast
- concat
- eltwise
- expand_dims
- matmul
- reduce
- reshape
- softmax
- squeeze
- stack
- stride_slice
Int8 operators
- conv_2d
- depthwsie_conv_2d
- eltwise
- mat_mul
- pooling
- softmax
- quantize
- dequantize
Tensorflow Keras Converter supported layers
- InputLayer
- Flatten
- Dense
- Conv2D
- MaxPooling2D
- Dropout
- DepthwiseConv2D
- Softmax
- BatchNormalization
- Activation
- ReLU
- Concatenate
- GlobalAveragePooling2D
- Add
- QuantizeLayer
- QuantizeWrapper