Tetap teratur dengan koleksi
Simpan dan kategorikan konten berdasarkan preferensi Anda.
aliran tensor:: operasi:: Dekuantisasi
#include <array_ops.h> Dekuantisasi tensor 'input' menjadi Tensor float.
Ringkasan
[min_range, max_range] adalah pelampung skalar yang menentukan rentang untuk data 'input'. Atribut 'mode' mengontrol penghitungan mana yang digunakan untuk mengonversi nilai float ke nilai terkuantisasinya.
Dalam mode 'MIN_COMBINED', setiap nilai tensor akan mengalami hal berikut:
if T == qint8: in[i] += (range(T) + 1)/ 2.0
out[i] = min_range + (in[i]* (max_range - min_range) / range(T))
di sini
range(T) = numeric_limits ::max() - numeric_limits ::min() range(T) = numeric_limits ::max() - numeric_limits ::min() range(T) = numeric_limits ::max() - numeric_limits ::min() Contoh Mode MIN_COMBINED
Jika masukan berasal dari QuantizedRelu6 , jenis keluarannya adalah quint8 (kisaran 0-255) tetapi kisaran QuantizedRelu6 yang mungkin adalah 0-6. Oleh karena itu, nilai min_range dan max_range adalah 0,0 dan 6,0. Dequantize pada quint8 akan mengambil setiap nilai, dilemparkan ke float, dan dikalikan dengan 6/255. Perhatikan bahwa jika quantizedtype adalah qint8, operasi ini juga akan menambahkan setiap nilai sebesar 128 sebelum casting.
Jika modenya adalah 'MIN_FIRST', maka pendekatan ini digunakan:
num_discrete_values = 1 << (# of bits in T)
range_adjust = num_discrete_values / (num_discrete_values - 1)
range = (range_max - range_min) * range_adjust
range_scale = range / num_discrete_values
const double offset_input = static_cast(input) - lowest_quantized;
result = range_min + ((input - numeric_limits::min()) * range_scale)
Contoh mode SKALA
Mode SCALED cocok dengan pendekatan kuantisasi yang digunakan dalam QuantizeAndDequantize{V2|V3} .
Jika modenya adalah SCALED , kita tidak menggunakan rentang penuh dari tipe keluaran, memilih untuk menghilangkan nilai simetri serendah mungkin (misalnya, rentang keluaran adalah -127 hingga 127, bukan -128 hingga 127 untuk kuantisasi 8 bit bertanda), sehingga 0,0 dipetakan ke 0.
Pertama-tama kita menemukan rentang nilai di tensor kita. Rentang yang kita gunakan selalu berpusat pada 0, jadi kita cari m sedemikian rupa
m = max(abs(input_min), abs(input_max))
Rentang tensor masukan kita adalah [-m, m] .
Selanjutnya, kita memilih keranjang kuantisasi titik tetap, [min_fixed, max_fixed] . Jika T ditandatangani, ini adalah
num_bits = sizeof(T) * 8
[min_fixed, max_fixed] =
[-(1 << (num_bits - 1) - 1), (1 << (num_bits - 1)) - 1] Sebaliknya, jika T tidak ditandatangani, rentang titik tetapnya adalah
[min_fixed, max_fixed] = [0, (1 << num_bits) - 1]
Dari sini kami menghitung faktor skala kami, s:
s = (2 * m) / (max_fixed - min_fixed)
Sekarang kita dapat melakukan dekuantisasi elemen tensor kita:
result = input * s
Argumen:
- ruang lingkup: Objek Lingkup
- min_range: Nilai skalar minimum yang mungkin dihasilkan untuk input.
- max_range: Nilai skalar maksimum yang mungkin dihasilkan untuk input.
Pengembalian:
Fungsi statis publik |
|---|
Mode (StringPiece x) | |
Atribut publik
Fungsi publik
simpul
::tensorflow::Node * node() const
operator::tensorflow::Input() const
operator::tensorflow::Keluaran
operator::tensorflow::Output() const
Fungsi statis publik
Mode
Attrs Mode(
StringPiece x
)
Kecuali dinyatakan lain, konten di halaman ini dilisensikan berdasarkan Lisensi Creative Commons Attribution 4.0, sedangkan contoh kode dilisensikan berdasarkan Lisensi Apache 2.0. Untuk mengetahui informasi selengkapnya, lihat Kebijakan Situs Google Developers. Java adalah merek dagang terdaftar dari Oracle dan/atau afiliasinya.
Terakhir diperbarui pada 2025-07-26 UTC.
[null,null,["Terakhir diperbarui pada 2025-07-26 UTC."],[],[],null,["# tensorflow::ops::Dequantize Class Reference\n\ntensorflow::ops::Dequantize\n===========================\n\n`#include \u003carray_ops.h\u003e`\n\n[Dequantize](/versions/r1.15/api_docs/cc/class/tensorflow/ops/dequantize#classtensorflow_1_1ops_1_1_dequantize) the 'input' tensor into a float [Tensor](/versions/r1.15/api_docs/cc/class/tensorflow/tensor#classtensorflow_1_1_tensor).\n\nSummary\n-------\n\n\\[min_range, max_range\\] are scalar floats that specify the range for the 'input' data. The 'mode' attribute controls exactly which calculations are used to convert the float values to their quantized equivalents.\n\nIn 'MIN_COMBINED' mode, each value of the tensor will undergo the following:\n\n\u003cbr /\u003e\n\n```transact-sql\nif T == qint8: in[i] += (range(T) + 1)/ 2.0\nout[i] = min_range + (in[i]* (max_range - min_range) / range(T))\n```\nhere `range(T) = numeric_limits`::max() - numeric_limits::min()\n\n\u003cbr /\u003e\n\n\n*MIN_COMBINED Mode Example*\n\nIf the input comes from a [QuantizedRelu6](/versions/r1.15/api_docs/cc/class/tensorflow/ops/quantized-relu6#classtensorflow_1_1ops_1_1_quantized_relu6), the output type is quint8 (range of 0-255) but the possible range of [QuantizedRelu6](/versions/r1.15/api_docs/cc/class/tensorflow/ops/quantized-relu6#classtensorflow_1_1ops_1_1_quantized_relu6) is 0-6. The min_range and max_range values are therefore 0.0 and 6.0. [Dequantize](/versions/r1.15/api_docs/cc/class/tensorflow/ops/dequantize#classtensorflow_1_1ops_1_1_dequantize) on quint8 will take each value, cast to float, and multiply by 6 / 255. Note that if quantizedtype is qint8, the operation will additionally add each value by 128 prior to casting.\n\nIf the mode is 'MIN_FIRST', then this approach is used:\n\n\n```gdscript\nnum_discrete_values = 1 \u003c\u003c (# of bits in T)\nrange_adjust = num_discrete_values / (num_discrete_values - 1)\nrange = (range_max - range_min) * range_adjust\nrange_scale = range / num_discrete_values\nconst double offset_input = static_cast(input) - lowest_quantized;\nresult = range_min + ((input - numeric_limits::min()) * range_scale)\n```\n\n\u003cbr /\u003e\n\n\n*SCALED mode Example*\n\n`SCALED` mode matches the quantization approach used in `QuantizeAndDequantize{V2|V3}`.\n\nIf the mode is `SCALED`, we do not use the full range of the output type, choosing to elide the lowest possible value for symmetry (e.g., output range is -127 to 127, not -128 to 127 for signed 8 bit quantization), so that 0.0 maps to 0.\n\nWe first find the range of values in our tensor. The range we use is always centered on 0, so we find m such that \n\n```scdoc\n m = max(abs(input_min), abs(input_max))\n```\n\n\u003cbr /\u003e\n\nOur input tensor range is then `[-m, m]`.\n\nNext, we choose our fixed-point quantization buckets, `[min_fixed, max_fixed]`. If T is signed, this is \n\n```scdoc\n num_bits = sizeof(T) * 8\n [min_fixed, max_fixed] =\n [-(1 \u003c\u003c (num_bits - 1) - 1), (1 \u003c\u003c (num_bits - 1)) - 1]\n```\n\n\u003cbr /\u003e\n\nOtherwise, if T is unsigned, the fixed-point range is \n\n```scdoc\n [min_fixed, max_fixed] = [0, (1 \u003c\u003c num_bits) - 1]\n```\n\n\u003cbr /\u003e\n\nFrom this we compute our scaling factor, s: \n\n```scdoc\n s = (2 * m) / (max_fixed - min_fixed)\n```\n\n\u003cbr /\u003e\n\nNow we can dequantize the elements of our tensor: \n\n```scdoc\nresult = input * s\n```\n\n\u003cbr /\u003e\n\nArguments:\n\n- scope: A [Scope](/versions/r1.15/api_docs/cc/class/tensorflow/scope#classtensorflow_1_1_scope) object\n- min_range: The minimum scalar value possibly produced for the input.\n- max_range: The maximum scalar value possibly produced for the input.\n\n\u003cbr /\u003e\n\nReturns:\n\n- [Output](/versions/r1.15/api_docs/cc/class/tensorflow/output#classtensorflow_1_1_output): The output tensor.\n\n\u003cbr /\u003e\n\n| ### Constructors and Destructors ||\n|---|---|\n| [Dequantize](#classtensorflow_1_1ops_1_1_dequantize_1ace6411557abc00c6e59649720be7d579)`(const ::`[tensorflow::Scope](/versions/r1.15/api_docs/cc/class/tensorflow/scope#classtensorflow_1_1_scope)` & scope, ::`[tensorflow::Input](/versions/r1.15/api_docs/cc/class/tensorflow/input#classtensorflow_1_1_input)` input, ::`[tensorflow::Input](/versions/r1.15/api_docs/cc/class/tensorflow/input#classtensorflow_1_1_input)` min_range, ::`[tensorflow::Input](/versions/r1.15/api_docs/cc/class/tensorflow/input#classtensorflow_1_1_input)` max_range)` ||\n| [Dequantize](#classtensorflow_1_1ops_1_1_dequantize_1afb71f46f9e4fc4922578ecd9116ad9b1)`(const ::`[tensorflow::Scope](/versions/r1.15/api_docs/cc/class/tensorflow/scope#classtensorflow_1_1_scope)` & scope, ::`[tensorflow::Input](/versions/r1.15/api_docs/cc/class/tensorflow/input#classtensorflow_1_1_input)` input, ::`[tensorflow::Input](/versions/r1.15/api_docs/cc/class/tensorflow/input#classtensorflow_1_1_input)` min_range, ::`[tensorflow::Input](/versions/r1.15/api_docs/cc/class/tensorflow/input#classtensorflow_1_1_input)` max_range, const `[Dequantize::Attrs](/versions/r1.15/api_docs/cc/struct/tensorflow/ops/dequantize/attrs#structtensorflow_1_1ops_1_1_dequantize_1_1_attrs)` & attrs)` ||\n\n| ### Public attributes ||\n|----------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------------------------|\n| [operation](#classtensorflow_1_1ops_1_1_dequantize_1a917ce29fbec6ef49406db9a374bde9aa) | [Operation](/versions/r1.15/api_docs/cc/class/tensorflow/operation#classtensorflow_1_1_operation) |\n| [output](#classtensorflow_1_1ops_1_1_dequantize_1a5c4618ae3d058bcd8547217612f8f41e) | `::`[tensorflow::Output](/versions/r1.15/api_docs/cc/class/tensorflow/output#classtensorflow_1_1_output) |\n\n| ### Public functions ||\n|----------------------------------------------------------------------------------------------------------------------|------------------------|\n| [node](#classtensorflow_1_1ops_1_1_dequantize_1a4bdeb613e4b88880638a67528cbd01f0)`() const ` | `::tensorflow::Node *` |\n| [operator::tensorflow::Input](#classtensorflow_1_1ops_1_1_dequantize_1ab1b62ee39a382d6e124eb62156c05525)`() const ` | ` ` ` ` |\n| [operator::tensorflow::Output](#classtensorflow_1_1ops_1_1_dequantize_1ae01ee2df9b62f7729848ca15ed70e8fc)`() const ` | ` ` ` ` |\n\n| ### Public static functions ||\n|----------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------------------------------|\n| [Mode](#classtensorflow_1_1ops_1_1_dequantize_1ac9873b34c5c0eb36296e0fe726644fc9)`(StringPiece x)` | [Attrs](/versions/r1.15/api_docs/cc/struct/tensorflow/ops/dequantize/attrs#structtensorflow_1_1ops_1_1_dequantize_1_1_attrs) |\n\n| ### Structs ||\n|----------------------------------------------------------------------------------------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------|\n| [tensorflow::ops::Dequantize::Attrs](/versions/r1.15/api_docs/cc/struct/tensorflow/ops/dequantize/attrs) | Optional attribute setters for [Dequantize](/versions/r1.15/api_docs/cc/class/tensorflow/ops/dequantize#classtensorflow_1_1ops_1_1_dequantize). |\n\nPublic attributes\n-----------------\n\n### operation\n\n```text\nOperation operation\n``` \n\n### output\n\n```text\n::tensorflow::Output output\n``` \n\nPublic functions\n----------------\n\n### Dequantize\n\n```gdscript\n Dequantize(\n const ::tensorflow::Scope & scope,\n ::tensorflow::Input input,\n ::tensorflow::Input min_range,\n ::tensorflow::Input max_range\n)\n``` \n\n### Dequantize\n\n```gdscript\n Dequantize(\n const ::tensorflow::Scope & scope,\n ::tensorflow::Input input,\n ::tensorflow::Input min_range,\n ::tensorflow::Input max_range,\n const Dequantize::Attrs & attrs\n)\n``` \n\n### node\n\n```gdscript\n::tensorflow::Node * node() const \n``` \n\n### operator::tensorflow::Input\n\n```gdscript\n operator::tensorflow::Input() const \n``` \n\n### operator::tensorflow::Output\n\n```gdscript\n operator::tensorflow::Output() const \n``` \n\nPublic static functions\n-----------------------\n\n### Mode\n\n```text\nAttrs Mode(\n StringPiece x\n)\n```"]]
