- Sort Score
- Result 10 results
- Languages All
Results 1 - 10 of 17 for quantized (0.39 sec)
-
tensorflow/compiler/mlir/quantization/tensorflow/passes/quantize_composite_functions.cc
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Thu Apr 25 16:01:03 UTC 2024 - 54.5K bytes - Viewed (0) -
tensorflow/compiler/mlir/lite/stablehlo/transforms/compose_uniform_quantized_type_pass.cc
}; // Matches the pattern for quantized convolution op and rewrites it to use // uniform quantized types. // // Currently assumes asymmetric per-tensor quantization for activations and // symmetric per-channel quantization for filters. // // This pattern represents the following derived equation, where: // * rn = real (expressed) value for tensor n // * qn = quantized value for tensor n // * sn = scale for tensor n
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Thu Apr 25 16:01:03 UTC 2024 - 64.6K bytes - Viewed (0) -
tensorflow/compiler/mlir/quantization/stablehlo/tests/passes/quantize_composite_functions.mlir
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Thu May 09 05:56:10 UTC 2024 - 91.6K bytes - Viewed (0) -
tensorflow/compiler/mlir/quantization/stablehlo/python/integration_test/quantize_model_test.py
if bias_fn: self.assertTrue(re.search('stablehlo.add.*xi32>', module_str)) # Consider if there is a way to check if activation fusion is properly # done in MLIR level. # Tests that the quantized graph outputs similar values. The rtol and atol # values are arbitrary. self.assertAllClose(new_outputs, expected_outputs, rtol=0.3, atol=0.2) # Due to other meta data, the compression is not exactly 1/4.
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Tue May 14 06:31:57 UTC 2024 - 51.4K bytes - Viewed (0) -
tensorflow/compiler/mlir/lite/quantization/lite/quantize_model_test.cc
Eq(TensorType_INT8)); // Verify FC bias should be int32 quantized. ASSERT_THAT(float_graph->tensors()->Get(float_op->inputs()->Get(2))->type(), Eq(TensorType_FLOAT32)); EXPECT_THAT(subgraph->tensors[op->inputs[2]].get()->type, Eq(TensorType_INT32)); // The output tensor of FC should be int8 quantized. ASSERT_THAT(float_graph->tensors()->Get(float_op->outputs()->Get(0))->type(),
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Wed Jun 12 23:15:24 UTC 2024 - 73.9K bytes - Viewed (0) -
tensorflow/compiler/mlir/tensorflow/transforms/lower_tf.cc
}; // This pass performs a manual conversion with FakeQuant, converting between // floating point and quantized space. It is designed to reproduce TF's // implementation, mirroring the previous XLA implementation. // // 1. Computing proper quantized bounds. This involves nudging the input bounds. // 2. Converting the input bounds to quantized space, rounding values. // 3. Convert back into floating point space.
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Thu Apr 25 16:01:03 UTC 2024 - 74.9K bytes - Viewed (0) -
tensorflow/compiler/mlir/quantization/tensorflow/python/integration_test/quantize_model_test_base.py
) def _is_quantized_function(self, func: function_pb2.FunctionDef) -> bool: """Determine whether a FunctionDef is quantized. Args: func: A FunctionDef object. Returns: True iff `func` is quantized. """ return func.signature.name.startswith('quantized_') def _is_composite_function(self, func: function_pb2.FunctionDef) -> bool:
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Thu Mar 21 08:51:46 UTC 2024 - 51.2K bytes - Viewed (0) -
tensorflow/compiler/mlir/quantization/tensorflow/tests/quantize_composit_functions_debugging.mlir
// RUN: tf-quant-opt %s -split-input-file -quant-insert-quantized-functions -quant-quantize-composite-functions | FileCheck --check-prefix=TF %s // RUN: tf-quant-opt %s -split-input-file -quant-insert-quantized-functions -quant-quantize-composite-functions='target-opset=XLA' | FileCheck --check-prefix=XLA %s
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Mon Nov 06 01:23:21 UTC 2023 - 80.5K bytes - Viewed (0) -
tensorflow/compiler/mlir/lite/flatbuffer_import.cc
// If the input `tensor` has scale/zero_point, `res` should have quantized // type, thus none stats op is required and nullptr is returned. // If the min max information is invalid, nullptr is returned. mlir::Operation* ConvertMinMaxToStatsOp(const TensorT& tensor, OpBuilder b, Value res) { // If the `tensor` has scale/zero_point, it must have been quantized, then the
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Tue May 21 18:21:50 UTC 2024 - 66.8K bytes - Viewed (0) -
tensorflow/compiler/mlir/g3doc/_includes/tf_passes.md
pruned using DCE. ### `-tf-lower-quantized` _Lowers ops that require quantized input or output._ This pass rewrites all ops that have at least one input or output that must be a quantized type to ops whose inputs and outputs allow non-quantized types. Examples of quantized types are TF_Qint8 or TF_Quint8. An example is TF_DequantizeOp, which converts a quantized type to a float.
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Wed Aug 02 02:26:39 UTC 2023 - 96.4K bytes - Viewed (0)