- Sort Score
- Result 10 results
- Languages All
Results 1 - 10 of 15 for weight_only (0.64 sec)
-
tensorflow/compiler/mlir/quantization/tensorflow/tests/quantize_composite_functions_weight_only.mlir
// RUN: tf-quant-opt %s -split-input-file -quant-insert-quantized-functions='quantization-method=weight_only target-opset=XLA' -quant-quantize-composite-functions='quantization-method=weight_only target-opset=XLA enable-per-channel-quantization=true' -symbol-dce...
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Mon Oct 30 06:52:55 UTC 2023 - 11.3K bytes - Viewed (0) -
tensorflow/compiler/mlir/quantization/tensorflow/tests/preprocess_op_weight_only.mlir
// RUN: tf-quant-opt %s -split-input-file -quant-preprocess-op='target-opset=XLA quantization-method=weight_only enable-per-channel-quantization=false' | FileCheck --check-prefix PerTensor %s // RUN: tf-quant-opt %s -split-input-file -quant-preprocess-op='target-opset=XLA quantization-method=weight_only enable-per-channel-quantization=true' | FileCheck --check-prefix PerChannel %s module { // For XLA weight-only per-channel depthwise convolution, tensor shape should have
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Mon Oct 30 06:52:55 UTC 2023 - 4.7K bytes - Viewed (0) -
tensorflow/compiler/mlir/quantization/tensorflow/tests/insert_quantized_functions_weight_only.mlir
// RUN: tf-quant-opt %s -quant-insert-quantized-functions='quantization-method=weight_only target-opset=XLA' | FileCheck %s // Empty module module { func.func @simple_fn(%arg0: tensor<*xf32>) -> tensor<*xf32> { func.return %arg0 : tensor<*xf32> } } // CHECK-NOT: func private @internal_dequantize_f32 // CHECK-NOT: func private @internal_conv3d_fn // CHECK-NOT: func private @internal_batch_matmul_fn
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Thu Feb 16 03:34:36 UTC 2023 - 843 bytes - Viewed (0) -
tensorflow/compiler/mlir/quantization/tensorflow/passes/lift_quantizable_spots_as_functions_drq.cc
clEnumValN(tensorflow::quantization::QuantizationMethod:: METHOD_STATIC_RANGE_WEIGHT_ONLY_INT8, "weight_only", "Post-training weight_only quantizaiton"))}; }; class CheckQuantizableOps : public mlir::OpRewritePattern<TF::PartitionedCallOp> { public: explicit CheckQuantizableOps(MLIRContext* context,
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Thu Apr 25 16:01:03 UTC 2024 - 8.5K bytes - Viewed (0) -
tensorflow/compiler/mlir/lite/quantization/tools/tflite_op_coverage_spec_getters_gen.cc
std::string dynamic_quant_kernel_support_regex = "bool GetDynamicRangeQuantKernelSupport() { return true; }"; raw_ostream &os = *ostream; std::vector<std::string> weight_only; llvm::sort(defs, LessRecord()); os.indent(0) << "const std::set<std::string> &ExportDynamicRangeSpec() {\n"; os.indent(2) << "static const std::set<std::string> * result =\n";
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Wed Sep 06 19:12:29 UTC 2023 - 12.7K bytes - Viewed (0) -
tensorflow/compiler/mlir/quantization/tensorflow/passes/insert_quantized_functions.cc
clEnumValN(tensorflow::quantization::QuantizationMethod:: METHOD_STATIC_RANGE_WEIGHT_ONLY_INT8, "weight_only", "Post-training weight_only quantizaiton"))}; Option<OpSet> op_set_{ *this, "target-opset", llvm::cl::init(OpSet::TF), llvm::cl::desc("Choose target opset."), llvm::cl::values(
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/stablehlo/utils/fill_quantization_options.cc
break; // Note: This is weight-only quantization by default, but with the legacy // flag "--force_dynamic_range_in_kernel", a DRQ behavior will be forced // in the kernel. case PresetQuantizationMethod::WEIGHT_ONLY: weight_component = custom_method.add_quantization_component_spec(); SetQuantizationComponentSpec(weight_component, QuantizationComponentSpec::COMPONENT_WEIGHT,
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Tue Mar 05 08:32:43 UTC 2024 - 6.2K bytes - Viewed (0) -
tensorflow/compiler/mlir/quantization/stablehlo/quantization_options.proto
// Apply default weight-only quantization. Weights are quantized during // conversion, then dequantized during inference. // Activation: f32, Weight: qi8, Bias: f32 WEIGHT_ONLY = 1; // Apply default dynamic range quantization. Quantized tensor value's // ranges are determined during graph runtime. // Activation: f32, Weight: qi8, Bias: f32
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Thu Jun 22 02:20:05 UTC 2023 - 3.6K bytes - Viewed (0) -
tensorflow/compiler/mlir/quantization/tensorflow/tests/lift_quantizable_spots_as_functions_drq.mlir
// CHECK-LABEL: private @composite_conv3d_fn_1 // WEIGHTONLY-DAG: %[[CST:.*]] = "tf.Const"() {{.*}} : () -> tensor<2x3x3x3x2xf32> // WEIGHTONLY: %[[PARTITIONEDCALL_0:.*]] = "tf.PartitionedCall"(%arg0, %[[CST]]) // WEIGHTONLY-SAME: f = @composite_conv3d_fn_1}> // WEIGHTONLY: {_tfl_quant_trait = "fully_quantizable" // WEIGHTONLY: %[[RELU:.*]] = "tf.Relu"(%[[PARTITIONEDCALL_0]]) // WEIGHTONLY: return %[[RELU]]
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Mon Oct 30 06:52:55 UTC 2023 - 11.8K bytes - Viewed (0) -
tensorflow/compiler/mlir/quantization/tensorflow/passes/preprocess_op.cc
"drq", "Post-training dynamic-range quantizaiton"), clEnumValN(tensorflow::quantization::QuantizationMethod:: METHOD_STATIC_RANGE_WEIGHT_ONLY_INT8, "weight_only", "Post-training weight-only quantizaiton"))}; Option<bool> enable_per_channel_quantization_{ *this, "enable-per-channel-quantization", llvm::cl::init(false),
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Thu Apr 25 16:01:03 UTC 2024 - 11.4K bytes - Viewed (0)