- Sort Score
- Result 10 results
- Languages All
Results 1 - 7 of 7 for quantize_i8 (0.23 sec)
-
tensorflow/compiler/mlir/quantization/tensorflow/tests/fake_quant_e2e_flow.mlir
// CHECK-NEXT: %[[V3:.*]] = "tf.PartitionedCall"(%[[V2]], %[[CST_4]], %[[CST_5]]) <{config = "", config_proto = "", executor_type = "", f = @dequantize_i8}> : (tensor<*xi8>, tensor<f32>, tensor<i32>) -> tensor<*xf32> // CHECK-NEXT: return %[[V3]] : tensor<*xf32> // CHECK: func private @quantize_i8( // CHECK: func private @dequantize_i8(
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Mon Oct 30 06:52:55 UTC 2023 - 3.5K bytes - Viewed (0) -
tensorflow/compiler/mlir/quantization/tensorflow/tests/insert_quantized_functions.mlir
// UQ-CHECK: func private @quantized_matmul_with_relu_fn // UQ-CHECK: func private @quantized_matmul_with_relu6_fn // UQ-CHECK: func private @quantize_i8 // UQ-CHECK: func private @quantize_i32
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Tue Aug 29 01:13:58 UTC 2023 - 3.3K bytes - Viewed (0) -
tensorflow/compiler/mlir/quantization/tensorflow/tests/replace_cast_hacks_with_tf_xla_ops_large_constants.mlir
%4 = "tf.PartitionedCall"(%3, %cst_2, %cst_3) {config = "", config_proto = "", executor_type = "", f = @dequantize_i8} : (tensor<1x2240x1120x512xi8>, tensor<f32>, tensor<i32>) -> tensor<1x2240x1120x512xf32> return %4 : tensor<1x2240x1120x512xf32> } func.func private @quantize_i8(%arg0: tensor<1x2240x2240x3xf32>, %arg1: tensor<f32>, %arg2: tensor<i32>) -> tensor<1x2240x2240x3xi8> {
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Mon Oct 30 06:52:55 UTC 2023 - 5.9K bytes - Viewed (0) -
tensorflow/compiler/mlir/quantization/stablehlo/tests/components/post_calibration_component.mlir
// CHECK-NO-UNPACK: %[[QUANTIZE_0:.+]] = stablehlo.uniform_quantize %[[ARG_0]] : (tensor<1x1024xf32>) -> tensor<1x1024x!quant.uniform<i8:f32, {{.*}}>> // CHECK-NO-UNPACK: %[[DOT:.+]] = stablehlo.dot_general %[[QUANTIZE_0]], %[[CONST]]
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Thu Apr 25 01:09:50 UTC 2024 - 6.7K bytes - Viewed (0) -
tensorflow/compiler/mlir/lite/transforms/modify_io_nodes.cc
quantize_op.setOperand(new_arg); } else { input_type.print(llvm::errs() << "Requested input type "); quantize_op.emitError(" Couldn't be modified to the requested type."); return failure(); } new_input_types[i] = arg_type; arg.dropAllUses(); if (quantize_op.use_empty()) { quantize_op.erase(); } } else {
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Thu Apr 25 16:01:03 UTC 2024 - 8.9K bytes - Viewed (0) -
tensorflow/compiler/mlir/quantization/tensorflow/passes/insert_quantized_functions.cc
METHOD_STATIC_RANGE_WEIGHT_ONLY_INT8) { // Uniform quantized opset is not supported for weight-only as inputs for // weight quantization are floats. And only dequantize_i8 is used from the // quantized function library. function_library_map = { {OpSet::TF, kQuantizedFunctionLibraryInMLIR}, {OpSet::XLA, kQuantizedFunctionLibraryInMLIR_XLA_WEIGHT_ONLY}}; } else {
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Fri Mar 22 05:52:39 UTC 2024 - 8.7K bytes - Viewed (0) -
tensorflow/compiler/mlir/quantization/tensorflow/passes/quantized_function_library_xla_weight_only.mlir
} : (tensor<*xf32>, tensor<*xf32>) -> tensor<*xf32> func.return %out : tensor<*xf32> } // Used for legacy weight-only func.func @dequantize_i8(%input : tensor<*xi8>, %scale : tensor<*xf32>, %zp : tensor<*xi32>) -> tensor<*xf32> { // Use identity op to avoid the weight being constant-folded. %identity = "tf.Identity"(%input) : (tensor<*xi8>) -> tensor<*xi8>
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Fri Mar 03 15:43:38 UTC 2023 - 7K bytes - Viewed (0)