wehub/tensorflow--tensorflow
1
0
Fork 0
Code Issues Pull Requests Actions Deployments Models Agent Notes Packages Projects Releases Wiki Activity
Files
8a852e4b4e25ba6dce753bedbed9f0ae4c4ec7ca
tensorflow--tensorflow/tensorflow/compiler/mlir/lite/tests/quantize-dynamic-range.mlir
T
wehub-resource-sync 8a852e4b4e
cffconvert / validate (push) Has been skipped
Details
License Check / license-check (push) Failing after 2s
Details
chore: import upstream snapshot with attribution
2026-07-13 12:14:16 +08:00

326 lines
24 KiB
MLIR
Raw Blame History

// Copyright 2026 The TensorFlow Authors. All Rights Reserved.
//
// Licensed 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.
// ==============================================================================
// RUN: litert-opt %s -tfl-prepare-quantize-dynamic-range -tfl-quantize="enable-dynamic-range-quantization=true" | FileCheck %s
// RUN: litert-opt %s -tfl-prepare-quantize-dynamic-range -tfl-quantize="enable-dynamic-range-quantization=true enable-weight-only-quantization=true" | FileCheck --check-prefix=PerChannelWeightOnly %s
// RUN: litert-opt %s -tfl-prepare-quantize-dynamic-range="enable-dynamic-range-per-channel-quantization=false" -tfl-quantize="enable-dynamic-range-quantization=true" | FileCheck --check-prefix=PerTensor %s
// RUN: litert-opt %s -tfl-prepare-quantize-dynamic-range="enable-dynamic-range-per-channel-quantization=false" -tfl-quantize="enable-dynamic-range-quantization=true enable-weight-only-quantization=true" | FileCheck --check-prefix=PerTensorWeightOnly %s
// RUN: litert-opt %s -tfl-prepare-quantize-dynamic-range="enable-dynamic-range-per-channel-quantization=false" -tfl-quantize="enable-dynamic-range-quantization=true ops-blocklist=tfl.conv_2d" | FileCheck --check-prefix=BLOCK %s
// RUN: litert-opt %s -tfl-prepare-quantize-dynamic-range="enable-custom-op-quantization=CustomTestOp=1" -tfl-quantize="enable-dynamic-range-quantization=true enable-custom-op-weight-only=CustomTestOp=true" | FileCheck --check-prefix=CustomOpWeightOnly %s
// RUN: litert-opt %s -tfl-prepare-quantize-dynamic-range="enable-custom-op-quantization=CustomTestOp=1" -tfl-quantize="enable-dynamic-range-quantization=true enable-custom-op-weight-only=CustomTestOp=false" | FileCheck --check-prefix=CustomOpNotWeightOnly %s
// CHECK-LABEL: QuantizeConv2D
// PerTensor-LABEL: QuantizeConv2D
// PerChannelWeightOnly-LABEL: QuantizeConv2D
// PerTensorWeightOnly-LABEL: QuantizeConv2D
// BLOCK-LABEL: QuantizeConv2D
func.func @QuantizeConv2D(%arg0: tensor<1x224x224x3xf32>) -> tensor<1x112x112x64xf32> {
%w = arith.constant dense<1.270000e+02> : tensor<64x3x3x3xf32>
%b = arith.constant dense<-1.23697901> : tensor<64xf32>
%conv = "tfl.conv_2d"(%arg0, %w, %b) {dilation_h_factor = 1 : i32, dilation_w_factor = 1 : i32, fused_activation_function = "NONE", padding = "SAME", stride_h = 2 : i32, stride_w = 2 : i32} : (tensor<1x224x224x3xf32>, tensor<64x3x3x3xf32>, tensor<64xf32>) -> tensor<1x112x112x64xf32>
func.return %conv : tensor<1x112x112x64xf32>
// CHECK: %[[b:.*]] = arith.constant dense<-1.23697901> : tensor<64xf32>
// CHECK: %[[w:.*]] = "tfl.pseudo_qconst"() <{qtype = tensor<64x3x3x3x!quant.uniform<i8<-127:127>:f32:0, {
// CHECK: %[[conv:.*]] = "tfl.conv_2d"(%arg0, %[[w]], %[[b]]) <{
// CHECK-NOT: asymmetric_quantize_inputs = true
// CHECK-SAME: dilation_h_factor = 1 : i32
// CHECK: return %[[conv:.*]]
// PerTensor: %[[b:.*]] = arith.constant dense<-1.23697901> : tensor<64xf32>
// PerTensor: %[[w:.*]] = "tfl.pseudo_qconst"() <{qtype = tensor<64x3x3x3x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>
// PerTensor: %[[conv:.*]] = "tfl.conv_2d"(%arg0, %[[w]], %[[b]]) <{
// PerTensor-NOT: asymmetric_quantize_inputs = true
// PerTensor-SAME: dilation_h_factor = 1 : i32
// PerTensor: return %[[conv:.*]]
// PerChannelWeightOnly: %[[b:.*]] = arith.constant dense<-1.23697901> : tensor<64xf32>
// PerChannelWeightOnly: %[[w:.*]] = "tfl.pseudo_qconst"() <{qtype = tensor<64x3x3x3x!quant.uniform<i8<-127:127>:f32:0, {
// PerChannelWeightOnly: %[[dq_w:.*]] = "tfl.dequantize"(%[[w]]) : (tensor<64x3x3x3x!quant.uniform<i8<-127:127>:f32:0, {
// PerChannelWeightOnly: %[[conv:.*]] = "tfl.conv_2d"(%arg0, %[[dq_w]], %[[b]]) <{
// PerChannelWeightOnly-NOT: asymmetric_quantize_inputs = true
// PerChannelWeightOnly-SAME: dilation_h_factor = 1 : i32
// PerChannelWeightOnly: return %[[conv:.*]]
// PerTensorWeightOnly: %[[b:.*]] = arith.constant dense<-1.23697901> : tensor<64xf32>
// PerTensorWeightOnly: %[[w:.*]] = "tfl.pseudo_qconst"() <{qtype = tensor<64x3x3x3x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>
// PerTensorWeightOnly: %[[dq_w:.*]] = "tfl.dequantize"(%[[w]]) : (tensor<64x3x3x3x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>
// PerTensorWeightOnly: %[[conv:.*]] = "tfl.conv_2d"(%arg0, %[[dq_w]], %[[b]]) <{
// PerTensorWeightOnly-NOT: asymmetric_quantize_inputs = true
// PerTensorWeightOnly-SAME: dilation_h_factor = 1 : i32
// PerTensorWeightOnly: return %[[conv:.*]]
// BLOCK: %[[b:.*]] = arith.constant dense<-1.23697901> : tensor<64xf32>
// BLOCK: %[[w:.*]] = "tfl.pseudo_qconst"() <{qtype = tensor<64x3x3x3x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>
// BLOCK: %[[dq_w:.*]] = "tfl.dequantize"(%[[w]]) : (tensor<64x3x3x3x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>
// BLOCK: %[[conv:.*]] = "tfl.conv_2d"(%arg0, %[[dq_w]], %[[b]]) <{
// BLOCK: return %[[conv:.*]]
}
// CHECK-LABEL: QuantizeDepthwiseConv2D
// PerTensor-LABEL: QuantizeDepthwiseConv2D
func.func @QuantizeDepthwiseConv2D(%arg0: tensor<1x224x224x3xf32>) -> tensor<1x112x112x64xf32> {
%w = arith.constant dense<127.0> : tensor<64x3x3x3xf32>
%b = arith.constant dense<0.0> : tensor<64xf32>
%dconv = "tfl.depthwise_conv_2d"(%arg0, %w, %b) {depth_multiplier = 4 : i32, dilation_h_factor = 1 : i32, dilation_w_factor = 1 : i32, fused_activation_function = "NONE", padding = "VALID", stride_h = 4 : i32, stride_w = 5 : i32} : (tensor<1x224x224x3xf32>, tensor<64x3x3x3xf32>, tensor<64xf32>) -> tensor<1x112x112x64xf32>
func.return %dconv : tensor<1x112x112x64xf32>
// CHECK: %[[b:.*]] = arith.constant dense<0.000000e+00> : tensor<64xf32>
// CHECK: %[[w:.*]] = "tfl.pseudo_qconst"() <{qtype = tensor<64x3x3x3x!quant.uniform<i8<-127:127>:f32:3, {1.000000e+00,1.000000e+00,1.000000e+00}
// CHECK: %[[dconv:.*]] = "tfl.depthwise_conv_2d"(%arg0, %[[w]], %[[b]]) <{
// CHECK-NOT: asymmetric_quantize_inputs = true
// CHECK-SAME: depth_multiplier = 4 : i32
// CHECK: return %[[dconv:.*]]
// PerTensor: %[[b:.*]] = arith.constant dense<0.000000e+00> : tensor<64xf32>
// PerTensor: %[[w:.*]] = "tfl.pseudo_qconst"() <{qtype = tensor<64x3x3x3x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>
// PerTensor: %[[dconv:.*]] = "tfl.depthwise_conv_2d"(%arg0, %[[w]], %[[b]]) <{
// PerTensor-NOT: asymmetric_quantize_inputs = true
// PerTensor-SAME: depth_multiplier = 4 : i32
// PerTensor: return %[[dconv:.*]]
}
// CHECK-LABEL: QuantizeFullyConnected
// PerTensor-LABEL: QuantizeFullyConnected
// PerChannelWeightOnly-LABEL: QuantizeFullyConnected
// PerTensorWeightOnly-LABEL: QuantizeFullyConnected
func.func @QuantizeFullyConnected(%arg0: tensor<1x224x224x3xf32>) -> tensor<1x112x112x512xf32> {
%w = arith.constant dense<127.0> : tensor<512x12xf32>
%b = arith.constant dense<0.0> : tensor<512xf32>
%fc = "tfl.fully_connected"(%arg0, %w, %b) {fused_activation_function = "NONE", keep_num_dims = false, weights_format = "DEFAULT"} : (tensor<1x224x224x3xf32>, tensor<512x12xf32>, tensor<512xf32>) -> tensor<1x112x112x512xf32>
func.return %fc : tensor<1x112x112x512xf32>
// CHECK: %[[b:.*]] = arith.constant dense<0.000000e+00> : tensor<512xf32>
// CHECK: %[[w:.*]] = "tfl.pseudo_qconst"() <{qtype = tensor<512x12x!quant.uniform<i8<-127:127>:f32:0, {1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00
// CHECK: %[[fc:.*]] = "tfl.fully_connected"(%arg0, %[[w]], %[[b]]) <{
// CHECK-NOT: fused_activation_function = "NONE",
// CHECK-SAME: asymmetric_quantize_inputs = true,
// CHECK: return %[[fc:.*]]
// PerTensor: %[[b:.*]] = arith.constant dense<0.000000e+00> : tensor<512xf32>
// PerTensor: %[[w:.*]] = "tfl.pseudo_qconst"() <{qtype = tensor<512x12x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>
// PerTensor: %[[fc:.*]] = "tfl.fully_connected"(%arg0, %[[w]], %[[b]]) <{
// PerTensor-NOT: fused_activation_function = "NONE",
// PerTensor-SAME: asymmetric_quantize_inputs = true,
// PerTensor: return %[[fc:.*]]
// PerChannelWeightOnly: %[[b:.*]] = arith.constant dense<0.000000e+00> : tensor<512xf32>
// PerChannelWeightOnly: %[[w:.*]] = "tfl.pseudo_qconst"() <{qtype = tensor<512x12x!quant.uniform<i8<-127:127>:f32:0, {1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00
// PerChannelWeightOnly: %[[dq_w:.*]] = "tfl.dequantize"(%[[w]]) : (tensor<512x12x!quant.uniform<i8<-127:127>:f32:0, {1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00
// PerChannelWeightOnly: %[[fc:.*]] = "tfl.fully_connected"(%arg0, %[[dq_w]], %[[b]]) <{
// PerChannelWeightOnly-NOT: fused_activation_function = "NONE",
// PerChannelWeightOnly-SAME: asymmetric_quantize_inputs = true,
// PerChannelWeightOnly: return %[[fc:.*]]
// PerTensorWeightOnly: %[[b:.*]] = arith.constant dense<0.000000e+00> : tensor<512xf32>
// PerTensorWeightOnly: %[[w:.*]] = "tfl.pseudo_qconst"() <{qtype = tensor<512x12x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>
// PerTensorWeightOnly: %[[dq_w:.*]] = "tfl.dequantize"(%[[w]]) : (tensor<512x12x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>
// PerTensorWeightOnly: %[[fc:.*]] = "tfl.fully_connected"(%arg0, %[[dq_w]], %[[b]]) <{
// PerTensorWeightOnly-NOT: fused_activation_function = "NONE",
// PerTensorWeightOnly-SAME: asymmetric_quantize_inputs = true,
// PerTensorWeightOnly: return %[[fc:.*]]
}
// CHECK-LABEL: QuantizeMatmulWithActConst
// PerTensor-LABEL: QuantizeMatmulWithActConst
func.func @QuantizeMatmulWithActConst(%arg0: tensor<1x3x3x512xf32>) -> tensor<1x3x3x12xf32> {
%w = arith.constant dense<127.0> : tensor<512x12xf32>
%mm = "tfl.batch_matmul"(%arg0, %w) {adj_x = false, adj_y = false} : (tensor<1x3x3x512xf32>, tensor<512x12xf32>) -> tensor<1x3x3x12xf32>
func.return %mm : tensor<1x3x3x12xf32>
// CHECK: %[[w:.*]] = "tfl.pseudo_qconst"() <{qtype = tensor<512x12x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>,
// CHECK: %[[mm:.*]] = "tfl.batch_matmul"(%arg0, %[[w]]) <{adj_x = false, adj_y = false
// CHECK-SAME: , asymmetric_quantize_inputs = true
// CHECK: return %[[mm:.*]]
// PerTensor: %[[w:.*]] = "tfl.pseudo_qconst"() <{qtype = tensor<512x12x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>,
// PerTensor: %[[mm:.*]] = "tfl.batch_matmul"(%arg0, %[[w]]) <{adj_x = false, adj_y = false
// PerTensor-SAME: , asymmetric_quantize_inputs = true
// PerTensor: return %[[mm:.*]]
}
// CHECK-LABEL: QuantizeTransposeConvWeightOnly
// PerTensor-LABEL: QuantizeTransposeConvWeightOnly
// PerChannelWeightOnly-LABEL: QuantizeTransposeConvWeightOnly
// PerTensorWeightOnly-LABEL: QuantizeTransposeConvWeightOnly
func.func @QuantizeTransposeConvWeightOnly(%arg0: tensor<32x4x4x128xf32>, %arg1: tensor<4xi32>) -> tensor<1x32x42x128xf32> {
%w = arith.constant dense<127.0> : tensor<1x32x42x128xf32>
%b = arith.constant dense<0.0> : tensor<1x32x42x128xf32>
%tconv = "tfl.transpose_conv"(%arg1, %w, %arg0, %b) {padding = "SAME", stride_h = 2 : i32, stride_w = 2 : i32, fused_activation_function = "NONE"} : (tensor<4xi32>, tensor<1x32x42x128xf32>, tensor<32x4x4x128xf32>, tensor<1x32x42x128xf32>) -> tensor<1x32x42x128xf32>
func.return %tconv : tensor<1x32x42x128xf32>
// CHECK: %[[b:.*]] = arith.constant dense<0.000000e+00> : tensor<1x32x42x128xf32>
// CHECK: %[[w:.*]] = "tfl.pseudo_qconst"() <{qtype = tensor<1x32x42x128x!quant.uniform<i8<-127:127>:f32:0, {1.000000e+00}>>
// CHECK: %[[tconv:.*]] = "tfl.transpose_conv"(%arg1, %[[w:.*]], %arg0, %[[b]]) <{
// CHECK-NOT: asymmetric_quantize_inputs = true
// CHECK-SAME: padding = "SAME"
// CHECK: return %[[tconv:.*]]
// PerTensor: %[[b:.*]] = arith.constant dense<0.000000e+00> : tensor<1x32x42x128xf32>
// PerTensor: %[[w:.*]] = "tfl.pseudo_qconst"() <{qtype = tensor<1x32x42x128x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>
// PerTensor: %[[tconv:.*]] = "tfl.transpose_conv"(%arg1, %[[w:.*]], %arg0, %[[b]]) <{
// PerTensor-NOT: asymmetric_quantize_inputs = true
// PerTensor-SAME: padding = "SAME"
// PerTensor: return %[[tconv:.*]]
// PerChannelWeightOnly: %[[b:.*]] = arith.constant dense<0.000000e+00> : tensor<1x32x42x128xf32>
// PerChannelWeightOnly: %[[w:.*]] = "tfl.pseudo_qconst"() <{qtype = tensor<1x32x42x128x!quant.uniform<i8<-127:127>:f32:0, {1.000000e+00}>>
// PerChannelWeightOnly: %[[tconv:.*]] = "tfl.transpose_conv"(%arg1, %[[w]], %arg0, %[[b]]) <{
// PerChannelWeightOnly-NOT: asymmetric_quantize_inputs = true
// PerChannelWeightOnly-SAME: padding = "SAME"
// PerChannelWeightOnly: return %[[tconv:.*]]
// PerTensorWeightOnly: %[[b:.*]] = arith.constant dense<0.000000e+00> : tensor<1x32x42x128xf32>
// PerTensorWeightOnly: %[[w:.*]] = "tfl.pseudo_qconst"() <{qtype = tensor<1x32x42x128x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>
// PerTensorWeightOnly: %[[tconv:.*]] = "tfl.transpose_conv"(%arg1, %[[w]], %arg0, %[[b]]) <{
// PerTensorWeightOnly-NOT: asymmetric_quantize_inputs = true
// PerTensorWeightOnly-SAME: padding = "SAME"
// PerTensorWeightOnly: return %[[tconv:.*]]
}
// CHECK-LABEL: QuantizeGatherWeightOnly
// PerTensor-LABEL: QuantizeGatherWeightOnly
func.func @QuantizeGatherWeightOnly(%arg0: tensor<3xi32>) -> tensor<3x3x3x3xf32> {
%w = arith.constant dense<1.270000e+02> : tensor<64x3x3x3xf32>
%emb = "tfl.gather"(%w, %arg0) {axis = 0 : i32, batch_dims = 0 : i32} : (tensor<64x3x3x3xf32>, tensor<3xi32>) -> tensor<3x3x3x3xf32>
%emb_s = "quantfork.stats"(%emb) {layerStats = dense<[0.000000e+00, 1.000000e+01]> : tensor<2xf32>} : (tensor<3x3x3x3xf32>) -> tensor<3x3x3x3xf32>
func.return %emb_s : tensor<3x3x3x3xf32>
// CHECK: %[[q_w:.*]] = "tfl.pseudo_qconst"() <{qtype = tensor<64x3x3x3x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>
// CHECK: %[[dq_w:.*]] = "tfl.dequantize"(%[[q_w]]) : (tensor<64x3x3x3x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>) -> tensor<64x3x3x3xf32>
// CHECK: %[[emb:.*]] = "tfl.gather"(%[[dq_w]], %arg0)
// CHECK: return %[[emb:.*]]
// PerTensor: %[[q_w:.*]] = "tfl.pseudo_qconst"() <{qtype = tensor<64x3x3x3x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>
// PerTensor: %[[dq_w:.*]] = "tfl.dequantize"(%[[q_w]]) : (tensor<64x3x3x3x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>) -> tensor<64x3x3x3xf32>
// PerTensor: %[[emb:.*]] = "tfl.gather"(%[[dq_w]], %arg0)
// PerTensor: return %[[emb:.*]]
}
// CHECK-LABEL: QuantizeCustomOp
// CustomOpWeightOnly-LABEL: QuantizeCustomOp
// CustomOpNotWeightOnly-LABEL: QuantizeCustomOp
func.func @QuantizeCustomOp(%arg0: tensor<1x1x1x1xf32>) -> tensor<*xf32> attributes {tf.entry_function = {inputs = "input", outputs = "custom_op"}} {
%0 = "quantfork.stats"(%arg0) {layerStats = dense<[0.000000e+00, 2.550000e+02]> : tensor<2xf32>} : (tensor<1x1x1x1xf32>) -> tensor<1x1x1x1xf32>
%w = arith.constant dense<127.0> : tensor<1024x1x1x1xf32>
%custom = "tfl.custom"(%0, %w) {custom_code = "CustomTestOp", custom_option = #tfl<const_bytes : "0x">} : (tensor<1x1x1x1xf32>, tensor<1024x1x1x1xf32>) -> tensor<*xf32>
func.return %custom : tensor<*xf32>
// CHECK: %[[w:.*]] = arith.constant dense<1.270000e+02> : tensor<1024x1x1x1xf32>
// CHECK: %[[custom:.*]] = "tfl.custom"(%arg0, %[[w:.*]]) <{custom_code = "CustomTestOp", custom_option = #tfl<const_bytes : "0x">}>
// CHECK: return %[[custom:.*]]
// CustomOpWeightOnly: %[[q_w:.*]] = "tfl.pseudo_qconst"() <{qtype = tensor<1024x1x1x1x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>
// CustomOpWeightOnly: %[[dq_w:.*]] = "tfl.dequantize"(%[[q_w:.*]]) : (tensor<1024x1x1x1x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>) -> tensor<1024x1x1x1xf32>
// CustomOpWeightOnly: %[[custom:.*]] = "tfl.custom"(%arg0, %[[dq_w:.*]]) <{custom_code = "CustomTestOp", custom_option = #tfl<const_bytes : "0x">}>
// CustomOpWeightOnly: return %[[custom:.*]]
// CustomOpNotWeightOnly: %[[q_w:.*]] = "tfl.pseudo_qconst"() <{qtype = tensor<1024x1x1x1x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>
// CustomOpNotWeightOnly: %[[custom:.*]] = "tfl.custom"(%arg0, %[[q_w:.*]]) <{custom_code = "CustomTestOp", custom_option = #tfl<const_bytes : "0x">}>
// CustomOpNotWeightOnly: return %[[custom:.*]]
}
// CHECK-LABEL: NotQuantizeConv3D
// PerTensor-LABEL: NotQuantizeConv3D
// PerChannelWeightOnly-LABEL: NotQuantizeConv3D
// PerTensorWeightOnly-LABEL: NotQuantizeConv3D
func.func @NotQuantizeConv3D(%arg0: tensor<1x32x32x32x8xf32>) -> tensor<1x32x32x32x16xf32> {
%w = arith.constant dense<127.0> : tensor<1x1x1x8x16xf32>
%b = arith.constant dense<0.0> : tensor<16xf32>
%conv_3d = "tfl.conv_3d"(%arg0, %w, %b) {dilation_d_factor = 1 : i32, dilation_h_factor = 1 : i32, dilation_w_factor = 1 : i32, fused_activation_function = "NONE", padding = "SAME", stride_d = 1 : i32, stride_h = 1 : i32, stride_w = 1 : i32} : (tensor<1x32x32x32x8xf32>, tensor<1x1x1x8x16xf32>, tensor<16xf32>) -> tensor<1x32x32x32x16xf32>
func.return %conv_3d : tensor<1x32x32x32x16xf32>
// CHECK-DAG: %[[w:.*]] = arith.constant dense<1.270000e+02> : tensor<1x1x1x8x16xf32>
// CHECK-DAG: %[[b:.*]] = arith.constant dense<0.000000e+00> : tensor<16xf32>
// CHECK: %[[conv_3d:.*]] = "tfl.conv_3d"(%arg0, %[[w]], %[[b]]) <{dilation_d_factor = 1 : i32, dilation_h_factor = 1 : i32, dilation_w_factor = 1 : i32, fused_activation_function = "NONE", padding = "SAME", stride_d = 1 : i32, stride_h = 1 : i32, stride_w = 1 : i32}>
// CHECK: return %[[conv_3d:.*]]
// PerTensor: %[[w:.*]] = arith.constant dense<1.270000e+02> : tensor<1x1x1x8x16xf32>
// PerTensor: %[[b:.*]] = arith.constant dense<0.000000e+00> : tensor<16xf32>
// PerTensor: %[[conv_3d:.*]] = "tfl.conv_3d"(%arg0, %[[w]], %[[b]]) <{dilation_d_factor = 1 : i32, dilation_h_factor = 1 : i32, dilation_w_factor = 1 : i32, fused_activation_function = "NONE", padding = "SAME", stride_d = 1 : i32, stride_h = 1 : i32, stride_w = 1 : i32}>
// PerTensor: return %[[conv_3d:.*]]
// PerChannelWeightOnly: %[[w:.*]] = arith.constant dense<1.270000e+02> : tensor<1x1x1x8x16xf32>
// PerChannelWeightOnly: %[[b:.*]] = arith.constant dense<0.000000e+00> : tensor<16xf32>
// PerChannelWeightOnly: %[[conv_3d:.*]] = "tfl.conv_3d"(%arg0, %[[w]], %[[b]]) <{dilation_d_factor = 1 : i32, dilation_h_factor = 1 : i32, dilation_w_factor = 1 : i32, fused_activation_function = "NONE", padding = "SAME", stride_d = 1 : i32, stride_h = 1 : i32, stride_w = 1 : i32}>
// PerChannelWeightOnly: return %[[conv_3d:.*]]
// PerTensorWeightOnly: %[[w:.*]] = arith.constant dense<1.270000e+02> : tensor<1x1x1x8x16xf32>
// PerTensorWeightOnly: %[[b:.*]] = arith.constant dense<0.000000e+00> : tensor<16xf32>
// PerTensorWeightOnly: %[[conv_3d:.*]] = "tfl.conv_3d"(%arg0, %[[w]], %[[b]]) <{dilation_d_factor = 1 : i32, dilation_h_factor = 1 : i32, dilation_w_factor = 1 : i32, fused_activation_function = "NONE", padding = "SAME", stride_d = 1 : i32, stride_h = 1 : i32, stride_w = 1 : i32}>
// PerTensorWeightOnly: return %[[conv_3d:.*]]
}
// CHECK-LABEL: QuantizeMultiUses
// PerTensor-LABEL: QuantizeMultiUses
// BLOCK-LABEL: QuantizeMultiUses
func.func @QuantizeMultiUses(%arg0: tensor<1x224x224x3xf32>, %arg1: tensor<3xi32>) -> (tensor<1x112x112x112xf32>, tensor<3x3x3x3xf32>) {
%w = arith.constant dense<1.270000e+02> : tensor<64x3x3x3xf32>
%b = arith.constant dense<-1.23697901> : tensor<64xf32>
%conv = "tfl.conv_2d"(%arg0, %w, %b) {dilation_h_factor = 1 : i32, dilation_w_factor = 1 : i32, fused_activation_function = "NONE", padding = "SAME", stride_h = 2 : i32, stride_w = 2 : i32} : (tensor<1x224x224x3xf32>, tensor<64x3x3x3xf32>, tensor<64xf32>) -> tensor<1x112x112x64xf32>
%dconv = "tfl.depthwise_conv_2d"(%arg0, %w, %b) {depth_multiplier = 4 : i32, dilation_h_factor = 1 : i32, dilation_w_factor = 1 : i32, fused_activation_function = "NONE", padding = "VALID", stride_h = 4 : i32, stride_w = 5 : i32} : (tensor<1x224x224x3xf32>, tensor<64x3x3x3xf32>, tensor<64xf32>) -> tensor<1x112x112x64xf32>
%emb = "tfl.gather"(%w, %arg1) {axis = 0 : i32, batch_dims = 0 : i32} : (tensor<64x3x3x3xf32>, tensor<3xi32>) -> tensor<3x3x3x3xf32>
%bmm = "tfl.batch_matmul"(%conv, %dconv) {adj_x = false, adj_y = true} : (tensor<1x112x112x64xf32>, tensor<1x112x112x64xf32>) -> tensor<1x112x112x112xf32>
func.return %bmm, %emb : tensor<1x112x112x112xf32>, tensor<3x3x3x3xf32>
// CHECK-DAG: %[[b:.*]] = arith.constant dense<-1.23697901> : tensor<64xf32>
// CHECK-DAG: %[[w1:.*]] = "tfl.pseudo_qconst"() <{qtype = tensor<64x3x3x3x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>
// CHECK-DAG: %[[dq_w1:.*]] = "tfl.dequantize"(%[[w1]]) : (tensor<64x3x3x3x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>) -> tensor<64x3x3x3xf32>
// CHECK-DAG: %[[w2:.*]] = "tfl.pseudo_qconst"() <{qtype = tensor<64x3x3x3x!quant.uniform<i8<-127:127>:f32:3, {1.000000e+00,1.000000e+00,1.000000e+00}>
// CHECK-DAG: %[[w3:.*]] = "tfl.pseudo_qconst"() <{qtype = tensor<64x3x3x3x!quant.uniform<i8<-127:127>:f32:0, {1.000000e+00,1.000000e+00,1.000000e+00
// CHECK: %[[conv:.*]] = "tfl.conv_2d"(%arg0, %[[w3]], %[[b]])
// CHECK: %[[dconv:.*]] = "tfl.depthwise_conv_2d"(%arg0, %[[w2]], %[[b]])
// CHECK: %[[emb:.*]] = "tfl.gather"(%[[dq_w1]], %arg1)
// CHECK: %[[bmm:.*]] = "tfl.batch_matmul"(%[[conv]], %[[dconv]]) <{adj_x = false, adj_y = true
// CHECK-NOT: , asymmetric_quantize_inputs = true
// CHECK-SAME: }
// CHECK: return %[[bmm:.*]], %[[emb:.*]]
// PerTensor: %[[b:.*]] = arith.constant dense<-1.23697901> : tensor<64xf32>
// PerTensor: %[[w1:.*]] = "tfl.pseudo_qconst"() <{qtype = tensor<64x3x3x3x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>
// PerTensor: %[[dq_w1:.*]] = "tfl.dequantize"(%[[w1]]) : (tensor<64x3x3x3x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>) -> tensor<64x3x3x3xf32>
// PerTensor: %[[conv:.*]] = "tfl.conv_2d"(%arg0, %[[w1]], %[[b]])
// PerTensor: %[[dconv:.*]] = "tfl.depthwise_conv_2d"(%arg0, %[[w1]], %[[b]])
// PerTensor: %[[emb:.*]] = "tfl.gather"(%[[dq_w1]], %arg1)
// PerTensor: %[[bmm:.*]] = "tfl.batch_matmul"(%[[conv]], %[[dconv]]) <{adj_x = false, adj_y = true
// PerTensor-NOT: , asymmetric_quantize_inputs = true
// PerTensor-SAME: }
// PerTensor: return %[[bmm:.*]], %[[emb:.*]]
// PerChannelWeightOnly-DAG: %[[b:.*]] = arith.constant dense<-1.23697901> : tensor<64xf32>
// PerChannelWeightOnly-DAG: %[[w1:.*]] = "tfl.pseudo_qconst"() <{qtype = tensor<64x3x3x3x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>
// PerChannelWeightOnly-DAG: %[[dq_w1:.*]] = "tfl.dequantize"(%[[w1]]) : (tensor<64x3x3x3x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>) -> tensor<64x3x3x3xf32>
// PerChannelWeightOnly-DAG: %[[w2:.*]] = "tfl.pseudo_qconst"() <{qtype = tensor<64x3x3x3x!quant.uniform<i8<-127:127>:f32:3, {1.000000e+00,1.000000e+00,1.000000e+00}
// PerChannelWeightOnly-DAG: %[[dq_w2:.*]] = "tfl.dequantize"(%[[w2]]) : (tensor<64x3x3x3x!quant.uniform<i8<-127:127>:f32:3, {1.000000e+00,1.000000e+00,1.000000e+00}>>) -> tensor<64x3x3x3xf32>
// PerChannelWeightOnly-DAG: %[[w3:.*]] = "tfl.pseudo_qconst"() <{qtype = tensor<64x3x3x3x!quant.uniform<i8<-127:127>:f32:0, {1.000000e+00,1.000000e+00,1.000000e+00
// PerChannelWeightOnly-DAG: %[[dq_w3:.*]] = "tfl.dequantize"(%[[w3]]) : (tensor<64x3x3x3x!quant.uniform<i8<-127:127>:f32:0, {1.000000e+00,1.000000e+00,1.000000e+00
// PerChannelWeightOnly: %[[conv:.*]] = "tfl.conv_2d"(%arg0, %[[dq_w3]], %[[b]])
// PerChannelWeightOnly: %[[dconv:.*]] = "tfl.depthwise_conv_2d"(%arg0, %[[dq_w2]], %[[b]])
// PerChannelWeightOnly: %[[emb:.*]] = "tfl.gather"(%[[dq_w1]], %arg1)
// PerChannelWeightOnly: %[[bmm:.*]] = "tfl.batch_matmul"(%[[conv]], %[[dconv]]) <{adj_x = false, adj_y = true
// PerChannelWeightOnly-NOT: , asymmetric_quantize_inputs = true
// PerChannelWeightOnly-SAME: }
// PerChannelWeightOnly: return %[[bmm:.*]], %[[emb:.*]]
// BLOCK: %[[b:.*]] = arith.constant dense<-1.23697901> : tensor<64xf32>
// BLOCK: %[[w1:.*]] = "tfl.pseudo_qconst"() <{qtype = tensor<64x3x3x3x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>
// BLOCK: %[[dq_w1:.*]] = "tfl.dequantize"(%[[w1]]) : (tensor<64x3x3x3x!quant.uniform<i8<-127:127>:f32, 1.000000e+00>>) -> tensor<64x3x3x3xf32>
// BLOCK: %[[conv:.*]] = "tfl.conv_2d"(%arg0, %[[dq_w1]], %[[b]])
// BLOCK: %[[dconv:.*]] = "tfl.depthwise_conv_2d"(%arg0, %[[w1]], %[[b]])
// BLOCK: %[[emb:.*]] = "tfl.gather"(%[[dq_w1]], %arg1)
// BLOCK: %[[bmm:.*]] = "tfl.batch_matmul"(%[[conv]], %[[dconv]]) <{adj_x = false, adj_y = true
// BLOCK: return %[[bmm:.*]], %[[emb:.*]]
}
Reference in New Issue View Git Blame Copy Permalink