ce2d218fc8
* Added support for quantized broadcast_to operator * Added tests for quantizing broadcast_to model |
||
|---|---|---|
| .. | ||
| add_with_const_input.bin | ||
| argmax.bin | ||
| broadcast_to.bin | ||
| concat.bin | ||
| custom_op.bin | ||
| fc_qat.bin | ||
| fc.bin | ||
| lstm_calibrated2.bin | ||
| lstm_calibrated.bin | ||
| lstm_quantized2.bin | ||
| lstm_quantized.bin | ||
| maximum.bin | ||
| minimum.bin | ||
| mixed16x8.bin | ||
| mixed.bin | ||
| multi_input_add_reshape.bin | ||
| pack.bin | ||
| quantized_with_gather.bin | ||
| README.md | ||
| single_avg_pool_min_minus_5_max_plus_5.bin | ||
| single_conv_no_bias.bin | ||
| single_conv_weights_min_0_max_plus_10.bin | ||
| single_conv_weights_min_minus_127_max_plus_127.bin | ||
| single_softmax_min_minus_5_max_plus_5.bin | ||
| split.bin | ||
| svdf_calibrated.bin | ||
| svdf_quantized.bin | ||
| transpose.bin | ||
| unidirectional_sequence_lstm_calibrated.bin | ||
| unidirectional_sequence_lstm_quantized.bin | ||
| unpack.bin | ||
| weight_shared_between_convs.bin | ||
Test models for testing quantization
This directory contains test models for testing quantization.
Models
single_conv_weights_min_0_max_plus_10.bin
A floating point model with single convolution where all weights are integers between [0, 10] weights are randomly distributed. It is not guaranteed that min max for weights are going to appear in each channel. All activations have min maxes and activations are in range [0,10].single_conv_weights_min_minus_127_max_plus_127.bin
A floating point model with a single convolution where weights of the model are all integers that lie in range[-127, 127]. The weights have been put in such a way that each channel has at least one weight as -127 and one weight as 127. The activations are all in range: [-128, 127]. This means all bias computations should result in 1.0 scale.single_softmax_min_minus_5_max_5.bin
A floating point model with a single softmax. The input tensor has min and max in range [-5, 5], not necessarily -5 or +5.single_avg_pool_input_min_minus_5_max_5.bin
A floating point model with a single average pool. The input tensor has min and max in range [-5, 5], not necessarily -5 or +5.weight_shared_between_convs.bin
A floating point model with two convs that have a use the same weight tensor.multi_input_add_reshape.bin
A floating point model with two inputs with an add followed by a reshape.quantized_with_gather.bin
A floating point model with an input with a gather, modeling a situation of mapping categorical input to embeddings.