<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
<html>
<head>
<!-- Generated by HsColour, http://code.haskell.org/~malcolm/hscolour/ -->
<title>src/TensorFlow/Ops.hs</title>
<link type='text/css' rel='stylesheet' href='hscolour.css' />
</head>
<body>
<pre><a name="line-1"></a><span class='hs-comment'>-- Copyright 2016 TensorFlow authors.</span>
<a name="line-2"></a><span class='hs-comment'>--</span>
<a name="line-3"></a><span class='hs-comment'>-- Licensed under the Apache License, Version 2.0 (the "License");</span>
<a name="line-4"></a><span class='hs-comment'>-- you may not use this file except in compliance with the License.</span>
<a name="line-5"></a><span class='hs-comment'>-- You may obtain a copy of the License at</span>
<a name="line-6"></a><span class='hs-comment'>--</span>
<a name="line-7"></a><span class='hs-comment'>-- <a href="http://www.apache.org/licenses/LICENSE-2.0">http://www.apache.org/licenses/LICENSE-2.0</a></span>
<a name="line-8"></a><span class='hs-comment'>--</span>
<a name="line-9"></a><span class='hs-comment'>-- Unless required by applicable law or agreed to in writing, software</span>
<a name="line-10"></a><span class='hs-comment'>-- distributed under the License is distributed on an "AS IS" BASIS,</span>
<a name="line-11"></a><span class='hs-comment'>-- WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.</span>
<a name="line-12"></a><span class='hs-comment'>-- See the License for the specific language governing permissions and</span>
<a name="line-13"></a><span class='hs-comment'>-- limitations under the License.</span>
<a name="line-14"></a>
<a name="line-15"></a><span class='hs-comment'>-- | This module contains definitions for some built-in TensorFlow operations.</span>
<a name="line-16"></a><span class='hs-comment'>--</span>
<a name="line-17"></a><span class='hs-comment'>-- Note that certain, "stateful" ops like 'variable' and 'assign' return a</span>
<a name="line-18"></a><span class='hs-comment'>-- 'Build' action (e.g., @Build (Tensor Ref a)@ instead of a pure value; the</span>
<a name="line-19"></a><span class='hs-comment'>-- returned 'Tensor's are always rendered in the current 'Build' context. This</span>
<a name="line-20"></a><span class='hs-comment'>-- approach helps us avoid problems with inlining or common subexpression</span>
<a name="line-21"></a><span class='hs-comment'>-- elimination, by writing</span>
<a name="line-22"></a><span class='hs-comment'>--</span>
<a name="line-23"></a><span class='hs-comment'>-- > do</span>
<a name="line-24"></a><span class='hs-comment'>-- > v <- variable []</span>
<a name="line-25"></a><span class='hs-comment'>-- > w <- assign v 3</span>
<a name="line-26"></a><span class='hs-comment'>-- > render $ w * w</span>
<a name="line-27"></a><span class='hs-comment'>--</span>
<a name="line-28"></a><span class='hs-comment'>-- instead of</span>
<a name="line-29"></a><span class='hs-comment'>--</span>
<a name="line-30"></a><span class='hs-comment'>-- > let</span>
<a name="line-31"></a><span class='hs-comment'>-- > v = variable []</span>
<a name="line-32"></a><span class='hs-comment'>-- > w = assign v 3</span>
<a name="line-33"></a><span class='hs-comment'>-- > in w * w</span>
<a name="line-34"></a><span class='hs-comment'>--</span>
<a name="line-35"></a><span class='hs-comment'>-- since the latter could be reasonably transformed by the compiler into (or</span>
<a name="line-36"></a><span class='hs-comment'>-- vice versa)</span>
<a name="line-37"></a><span class='hs-comment'>--</span>
<a name="line-38"></a><span class='hs-comment'>-- > let</span>
<a name="line-39"></a><span class='hs-comment'>-- > v = variable []</span>
<a name="line-40"></a><span class='hs-comment'>-- > w = assign v 3</span>
<a name="line-41"></a><span class='hs-comment'>-- > w' = assign v 3</span>
<a name="line-42"></a><span class='hs-comment'>-- > in w * w'</span>
<a name="line-43"></a><span class='hs-comment'>--</span>
<a name="line-44"></a><span class='hs-comment'>-- Ops should return a 'Build' action if their original 'OpDef' marks them as</span>
<a name="line-45"></a><span class='hs-comment'>-- stateful, or if they take any Refs as input. (This mirrors the rules that</span>
<a name="line-46"></a><span class='hs-comment'>-- TensorFlow uses to avoid common subexpression elimination.)</span>
<a name="line-47"></a><span class='hs-comment'>{-# LANGUAGE ConstraintKinds #-}</span>
<a name="line-48"></a><span class='hs-comment'>{-# LANGUAGE DataKinds #-}</span>
<a name="line-49"></a><span class='hs-comment'>{-# LANGUAGE FlexibleInstances #-}</span>
<a name="line-50"></a><span class='hs-comment'>{-# LANGUAGE OverloadedLists #-}</span>
<a name="line-51"></a><span class='hs-comment'>{-# LANGUAGE OverloadedStrings #-}</span>
<a name="line-52"></a><span class='hs-comment'>{-# LANGUAGE RankNTypes #-}</span>
<a name="line-53"></a><span class='hs-comment'>{-# LANGUAGE ScopedTypeVariables #-}</span>
<a name="line-54"></a><span class='hs-comment'>{-# LANGUAGE TypeFamilies #-}</span>
<a name="line-55"></a><span class='hs-comment'>{-# LANGUAGE UndecidableInstances #-}</span>
<a name="line-56"></a><span class='hs-comment'>{-# OPTIONS_GHC -fno-warn-orphans #-}</span>
<a name="line-57"></a>
<a name="line-58"></a><span class='hs-keyword'>module</span> <span class='hs-conid'>TensorFlow</span><span class='hs-varop'>.</span><span class='hs-conid'>Ops</span>
<a name="line-59"></a> <span class='hs-layout'>(</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>add</span>
<a name="line-60"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>abs</span>
<a name="line-61"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>addN</span>
<a name="line-62"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>argMax</span>
<a name="line-63"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>assign</span>
<a name="line-64"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>broadcastGradientArgs</span>
<a name="line-65"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>cast</span>
<a name="line-66"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>concat</span>
<a name="line-67"></a> <span class='hs-layout'>,</span> <span class='hs-varid'>constant</span>
<a name="line-68"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>equal</span>
<a name="line-69"></a> <span class='hs-layout'>,</span> <span class='hs-varid'>expandDims</span>
<a name="line-70"></a> <span class='hs-layout'>,</span> <span class='hs-varid'>initializedVariable</span>
<a name="line-71"></a> <span class='hs-layout'>,</span> <span class='hs-varid'>zeroInitializedVariable</span>
<a name="line-72"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>fill</span>
<a name="line-73"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>oneHot</span>
<a name="line-74"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>matMul</span>
<a name="line-75"></a> <span class='hs-layout'>,</span> <span class='hs-varid'>matTranspose</span>
<a name="line-76"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>mean</span>
<a name="line-77"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>mul</span>
<a name="line-78"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>neg</span>
<a name="line-79"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>pack</span>
<a name="line-80"></a> <span class='hs-layout'>,</span> <span class='hs-varid'>placeholder</span>
<a name="line-81"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>range</span>
<a name="line-82"></a> <span class='hs-layout'>,</span> <span class='hs-varid'>reducedShape</span>
<a name="line-83"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>relu</span>
<a name="line-84"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>reluGrad</span>
<a name="line-85"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>reshape</span>
<a name="line-86"></a> <span class='hs-layout'>,</span> <span class='hs-varid'>restore</span>
<a name="line-87"></a> <span class='hs-layout'>,</span> <span class='hs-varid'>restoreFromName</span>
<a name="line-88"></a> <span class='hs-layout'>,</span> <span class='hs-varid'>save</span>
<a name="line-89"></a> <span class='hs-layout'>,</span> <span class='hs-varid'>scalar</span>
<a name="line-90"></a> <span class='hs-layout'>,</span> <span class='hs-varid'>shape</span>
<a name="line-91"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>sign</span>
<a name="line-92"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>size</span>
<a name="line-93"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>softmax</span>
<a name="line-94"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>softmaxCrossEntropyWithLogits</span>
<a name="line-95"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>sparseToDense</span>
<a name="line-96"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>sub</span>
<a name="line-97"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>sum</span>
<a name="line-98"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>transpose</span>
<a name="line-99"></a> <span class='hs-layout'>,</span> <span class='hs-varid'>truncatedNormal</span>
<a name="line-100"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>variable</span>
<a name="line-101"></a> <span class='hs-layout'>,</span> <span class='hs-varid'>vector</span>
<a name="line-102"></a> <span class='hs-layout'>,</span> <span class='hs-varid'>zeros</span>
<a name="line-103"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>zerosLike</span>
<a name="line-104"></a> <span class='hs-layout'>,</span> <span class='hs-varid'>scalarize</span>
<a name="line-105"></a> <span class='hs-layout'>)</span> <span class='hs-keyword'>where</span>
<a name="line-106"></a>
<a name="line-107"></a><span class='hs-keyword'>import</span> <span class='hs-conid'>Data</span><span class='hs-varop'>.</span><span class='hs-conid'>ByteString</span> <span class='hs-layout'>(</span><span class='hs-conid'>ByteString</span><span class='hs-layout'>)</span>
<a name="line-108"></a><span class='hs-keyword'>import</span> <span class='hs-conid'>Data</span><span class='hs-varop'>.</span><span class='hs-conid'>Complex</span> <span class='hs-layout'>(</span><span class='hs-conid'>Complex</span><span class='hs-layout'>)</span>
<a name="line-109"></a><span class='hs-keyword'>import</span> <span class='hs-conid'>Data</span><span class='hs-varop'>.</span><span class='hs-conid'>Int</span> <span class='hs-layout'>(</span><span class='hs-conid'>Int32</span><span class='hs-layout'>,</span> <span class='hs-conid'>Int64</span><span class='hs-layout'>)</span>
<a name="line-110"></a><span class='hs-keyword'>import</span> <span class='hs-conid'>Prelude</span> <span class='hs-varid'>hiding</span> <span class='hs-layout'>(</span><span class='hs-varid'>abs</span><span class='hs-layout'>,</span> <span class='hs-varid'>sum</span><span class='hs-layout'>,</span> <span class='hs-varid'>concat</span><span class='hs-layout'>)</span>
<a name="line-111"></a><span class='hs-keyword'>import</span> <span class='hs-conid'>Data</span><span class='hs-varop'>.</span><span class='hs-conid'>ProtoLens</span> <span class='hs-layout'>(</span><span class='hs-varid'>def</span><span class='hs-layout'>)</span>
<a name="line-112"></a><span class='hs-keyword'>import</span> <span class='hs-conid'>Data</span><span class='hs-varop'>.</span><span class='hs-conid'>Text</span><span class='hs-varop'>.</span><span class='hs-conid'>Encoding</span> <span class='hs-layout'>(</span><span class='hs-varid'>encodeUtf8</span><span class='hs-layout'>)</span>
<a name="line-113"></a><span class='hs-keyword'>import</span> <span class='hs-conid'>Lens</span><span class='hs-varop'>.</span><span class='hs-conid'>Family2</span> <span class='hs-layout'>(</span><span class='hs-layout'>(</span><span class='hs-varop'>.~</span><span class='hs-layout'>)</span><span class='hs-layout'>,</span> <span class='hs-layout'>(</span><span class='hs-varop'>&</span><span class='hs-layout'>)</span><span class='hs-layout'>)</span>
<a name="line-114"></a><span class='hs-keyword'>import</span> <span class='hs-conid'>Text</span><span class='hs-varop'>.</span><span class='hs-conid'>Printf</span> <span class='hs-layout'>(</span><span class='hs-varid'>printf</span><span class='hs-layout'>)</span>
<a name="line-115"></a><span class='hs-keyword'>import</span> <span class='hs-conid'>Proto</span><span class='hs-varop'>.</span><span class='hs-conid'>Tensorflow</span><span class='hs-varop'>.</span><span class='hs-conid'>Core</span><span class='hs-varop'>.</span><span class='hs-conid'>Framework</span><span class='hs-varop'>.</span><span class='hs-conid'>Tensor</span>
<a name="line-116"></a> <span class='hs-layout'>(</span> <span class='hs-conid'>TensorProto</span>
<a name="line-117"></a> <span class='hs-layout'>,</span> <span class='hs-varid'>dtype</span>
<a name="line-118"></a> <span class='hs-layout'>,</span> <span class='hs-varid'>tensorShape</span>
<a name="line-119"></a> <span class='hs-layout'>)</span>
<a name="line-120"></a><span class='hs-keyword'>import</span> <span class='hs-keyword'>qualified</span> <span class='hs-conid'>Proto</span><span class='hs-varop'>.</span><span class='hs-conid'>Tensorflow</span><span class='hs-varop'>.</span><span class='hs-conid'>Core</span><span class='hs-varop'>.</span><span class='hs-conid'>Framework</span><span class='hs-varop'>.</span><span class='hs-conid'>TensorShape</span>
<a name="line-121"></a> <span class='hs-keyword'>as</span> <span class='hs-conid'>TensorShape</span>
<a name="line-122"></a><span class='hs-keyword'>import</span> <span class='hs-conid'>TensorFlow</span><span class='hs-varop'>.</span><span class='hs-conid'>Build</span>
<a name="line-123"></a><span class='hs-keyword'>import</span> <span class='hs-conid'>TensorFlow</span><span class='hs-varop'>.</span><span class='hs-conid'>BuildOp</span>
<a name="line-124"></a><span class='hs-keyword'>import</span> <span class='hs-conid'>TensorFlow</span><span class='hs-varop'>.</span><span class='hs-conid'>ControlFlow</span> <span class='hs-layout'>(</span><span class='hs-varid'>group</span><span class='hs-layout'>)</span>
<a name="line-125"></a><span class='hs-keyword'>import</span> <span class='hs-conid'>TensorFlow</span><span class='hs-varop'>.</span><span class='hs-conid'>Output</span> <span class='hs-layout'>(</span><span class='hs-varid'>unNodeName</span><span class='hs-layout'>)</span>
<a name="line-126"></a><span class='hs-keyword'>import</span> <span class='hs-conid'>TensorFlow</span><span class='hs-varop'>.</span><span class='hs-conid'>Tensor</span>
<a name="line-127"></a><span class='hs-keyword'>import</span> <span class='hs-conid'>TensorFlow</span><span class='hs-varop'>.</span><span class='hs-conid'>Types</span>
<a name="line-128"></a>
<a name="line-129"></a><span class='hs-keyword'>import</span> <span class='hs-keyword'>qualified</span> <span class='hs-conid'>TensorFlow</span><span class='hs-varop'>.</span><span class='hs-conid'>GenOps</span><span class='hs-varop'>.</span><span class='hs-conid'>Core</span> <span class='hs-keyword'>as</span> <span class='hs-conid'>CoreOps</span>
<a name="line-130"></a>
<a name="line-131"></a><span class='hs-keyword'>import</span> <span class='hs-keyword'>qualified</span> <span class='hs-conid'>Prelude</span> <span class='hs-layout'>(</span><span class='hs-varid'>abs</span><span class='hs-layout'>)</span>
<a name="line-132"></a>
<a name="line-133"></a><a name="instance%20Num%20(Tensor%20v%20a)"></a><span class='hs-comment'>-- TODO: Look into hs-boot refactoring to allow mutually recursive imports.</span>
<a name="line-134"></a><a name="instance%20Num%20(Tensor%20v%20a)"></a><span class='hs-comment'>-- | Must be defined as an orphan because of the dependency order between Ops</span>
<a name="line-135"></a><a name="instance%20Num%20(Tensor%20v%20a)"></a><span class='hs-comment'>-- and Tensor.</span>
<a name="line-136"></a><a name="instance%20Num%20(Tensor%20v%20a)"></a><span class='hs-comment'>--</span>
<a name="line-137"></a><a name="instance%20Num%20(Tensor%20v%20a)"></a><span class='hs-comment'>-- The indirect constraint "v ~ Value" helps disambiguate types, for example in</span>
<a name="line-138"></a><a name="instance%20Num%20(Tensor%20v%20a)"></a><span class='hs-comment'>-- "neg 1 :: Tensor Value Float", it helps find the type of the subexpression</span>
<a name="line-139"></a><a name="instance%20Num%20(Tensor%20v%20a)"></a><span class='hs-comment'>-- "1".</span>
<a name="line-140"></a><a name="instance%20Num%20(Tensor%20v%20a)"></a><span class='hs-keyword'>instance</span> <span class='hs-layout'>(</span> <span class='hs-conid'>TensorType</span> <span class='hs-varid'>a</span>
<a name="line-141"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>Num</span> <span class='hs-varid'>a</span>
<a name="line-142"></a> <span class='hs-layout'>,</span> <span class='hs-varid'>v</span> <span class='hs-keyglyph'>~</span> <span class='hs-conid'>Value</span>
<a name="line-143"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>OneOf</span> <span class='hs-chr'>'</span><span class='hs-keyglyph'>[</span> <span class='hs-conid'>Double</span><span class='hs-layout'>,</span> <span class='hs-conid'>Float</span><span class='hs-layout'>,</span> <span class='hs-conid'>Int32</span><span class='hs-layout'>,</span> <span class='hs-conid'>Int64</span>
<a name="line-144"></a> <span class='hs-layout'>,</span> <span class='hs-conid'>Complex</span> <span class='hs-conid'>Float</span><span class='hs-layout'>,</span> <span class='hs-conid'>Complex</span> <span class='hs-conid'>Double</span><span class='hs-keyglyph'>]</span> <span class='hs-varid'>a</span><span class='hs-layout'>)</span> <span class='hs-keyglyph'>=></span> <span class='hs-conid'>Num</span> <span class='hs-layout'>(</span><span class='hs-conid'>Tensor</span> <span class='hs-varid'>v</span> <span class='hs-varid'>a</span><span class='hs-layout'>)</span> <span class='hs-keyword'>where</span>
<a name="line-145"></a> <span class='hs-layout'>(</span><span class='hs-varop'>+</span><span class='hs-layout'>)</span> <span class='hs-keyglyph'>=</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>add</span>
<a name="line-146"></a> <span class='hs-layout'>(</span><span class='hs-varop'>*</span><span class='hs-layout'>)</span> <span class='hs-keyglyph'>=</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>mul</span>
<a name="line-147"></a> <span class='hs-layout'>(</span><span class='hs-comment'>-</span><span class='hs-layout'>)</span> <span class='hs-keyglyph'>=</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>sub</span>
<a name="line-148"></a> <span class='hs-varid'>abs</span> <span class='hs-keyglyph'>=</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>abs</span>
<a name="line-149"></a> <span class='hs-varid'>fromInteger</span> <span class='hs-keyglyph'>=</span> <span class='hs-varid'>scalar</span> <span class='hs-varop'>.</span> <span class='hs-varid'>fromInteger</span>
<a name="line-150"></a> <span class='hs-varid'>signum</span> <span class='hs-keyglyph'>=</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>sign</span>
<a name="line-151"></a> <span class='hs-varid'>negate</span> <span class='hs-keyglyph'>=</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>neg</span>
<a name="line-152"></a>
<a name="line-153"></a><a name="matTranspose"></a><span class='hs-definition'>matTranspose</span> <span class='hs-keyglyph'>::</span> <span class='hs-keyword'>forall</span> <span class='hs-varid'>a</span> <span class='hs-varid'>v</span> <span class='hs-varop'>.</span> <span class='hs-conid'>TensorType</span> <span class='hs-varid'>a</span>
<a name="line-154"></a> <span class='hs-keyglyph'>=></span> <span class='hs-conid'>Tensor</span> <span class='hs-varid'>v</span> <span class='hs-varid'>a</span> <span class='hs-keyglyph'>-></span> <span class='hs-conid'>Tensor</span> <span class='hs-conid'>Value</span> <span class='hs-varid'>a</span>
<a name="line-155"></a><span class='hs-definition'>matTranspose</span> <span class='hs-keyglyph'>=</span> <span class='hs-varid'>flip</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>transpose</span> <span class='hs-layout'>(</span><span class='hs-varid'>vector</span> <span class='hs-keyglyph'>[</span><span class='hs-num'>1</span><span class='hs-layout'>,</span> <span class='hs-num'>0</span> <span class='hs-keyglyph'>::</span> <span class='hs-conid'>Int32</span><span class='hs-keyglyph'>]</span><span class='hs-layout'>)</span>
<a name="line-156"></a>
<a name="line-157"></a><a name="placeholder"></a><span class='hs-definition'>placeholder</span> <span class='hs-keyglyph'>::</span> <span class='hs-keyword'>forall</span> <span class='hs-varid'>a</span> <span class='hs-varop'>.</span> <span class='hs-conid'>TensorType</span> <span class='hs-varid'>a</span> <span class='hs-keyglyph'>=></span> <span class='hs-conid'>Shape</span> <span class='hs-keyglyph'>-></span> <span class='hs-conid'>Build</span> <span class='hs-layout'>(</span><span class='hs-conid'>Tensor</span> <span class='hs-conid'>Value</span> <span class='hs-varid'>a</span><span class='hs-layout'>)</span>
<a name="line-158"></a><span class='hs-definition'>placeholder</span> <span class='hs-varid'>shape'</span> <span class='hs-keyglyph'>=</span>
<a name="line-159"></a> <span class='hs-varid'>buildOp</span> <span class='hs-varop'>$</span> <span class='hs-varid'>opDef</span> <span class='hs-str'>"Placeholder"</span>
<a name="line-160"></a> <span class='hs-varop'>&</span> <span class='hs-varid'>opAttr</span> <span class='hs-str'>"dtype"</span> <span class='hs-varop'>.~</span> <span class='hs-varid'>tensorType</span> <span class='hs-layout'>(</span><span class='hs-varid'>undefined</span> <span class='hs-keyglyph'>::</span> <span class='hs-varid'>a</span><span class='hs-layout'>)</span>
<a name="line-161"></a> <span class='hs-varop'>&</span> <span class='hs-varid'>opAttr</span> <span class='hs-str'>"shape"</span> <span class='hs-varop'>.~</span> <span class='hs-varid'>shape'</span>
<a name="line-162"></a>
<a name="line-163"></a><a name="initializedVariable"></a><span class='hs-comment'>-- | Creates a variable initialized to the given value.</span>
<a name="line-164"></a><span class='hs-comment'>-- Initialization happens next time session runs.</span>
<a name="line-165"></a><span class='hs-definition'>initializedVariable</span> <span class='hs-keyglyph'>::</span> <span class='hs-keyword'>forall</span> <span class='hs-varid'>a</span> <span class='hs-varop'>.</span> <span class='hs-conid'>TensorType</span> <span class='hs-varid'>a</span>
<a name="line-166"></a> <span class='hs-keyglyph'>=></span> <span class='hs-conid'>Tensor</span> <span class='hs-conid'>Value</span> <span class='hs-varid'>a</span> <span class='hs-keyglyph'>-></span> <span class='hs-conid'>Build</span> <span class='hs-layout'>(</span><span class='hs-conid'>Tensor</span> <span class='hs-conid'>Ref</span> <span class='hs-varid'>a</span><span class='hs-layout'>)</span>
<a name="line-167"></a><span class='hs-definition'>initializedVariable</span> <span class='hs-varid'>initializer</span> <span class='hs-keyglyph'>=</span> <span class='hs-keyword'>do</span>
<a name="line-168"></a> <span class='hs-varid'>v</span> <span class='hs-keyglyph'><-</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>variable</span> <span class='hs-conid'>[]</span> <span class='hs-comment'>-- The shape is not known initially.</span>
<a name="line-169"></a> <span class='hs-layout'>(</span><span class='hs-varid'>i</span> <span class='hs-keyglyph'>::</span> <span class='hs-conid'>Tensor</span> <span class='hs-conid'>Ref</span> <span class='hs-varid'>a</span><span class='hs-layout'>)</span> <span class='hs-keyglyph'><-</span>
<a name="line-170"></a> <span class='hs-varid'>buildOp</span> <span class='hs-layout'>(</span><span class='hs-varid'>opDef</span> <span class='hs-str'>"Assign"</span>
<a name="line-171"></a> <span class='hs-varop'>&</span> <span class='hs-varid'>opAttr</span> <span class='hs-str'>"T"</span> <span class='hs-varop'>.~</span> <span class='hs-varid'>tensorType</span> <span class='hs-layout'>(</span><span class='hs-varid'>undefined</span> <span class='hs-keyglyph'>::</span> <span class='hs-varid'>a</span><span class='hs-layout'>)</span>
<a name="line-172"></a> <span class='hs-varop'>&</span> <span class='hs-varid'>opAttr</span> <span class='hs-str'>"use_locking"</span> <span class='hs-varop'>.~</span> <span class='hs-conid'>True</span>
<a name="line-173"></a> <span class='hs-varop'>&</span> <span class='hs-varid'>opAttr</span> <span class='hs-str'>"validate_shape"</span> <span class='hs-varop'>.~</span> <span class='hs-conid'>False</span>
<a name="line-174"></a> <span class='hs-layout'>)</span>
<a name="line-175"></a> <span class='hs-varid'>v</span> <span class='hs-varid'>initializer</span>
<a name="line-176"></a> <span class='hs-varid'>addInitializer</span> <span class='hs-varop'>=<<</span> <span class='hs-varid'>group</span> <span class='hs-varid'>i</span>
<a name="line-177"></a> <span class='hs-varid'>return</span> <span class='hs-varid'>v</span>
<a name="line-178"></a>
<a name="line-179"></a><a name="zeroInitializedVariable"></a><span class='hs-comment'>-- | Creates a zero-initialized variable with the given shape.</span>
<a name="line-180"></a><span class='hs-definition'>zeroInitializedVariable</span>
<a name="line-181"></a> <span class='hs-keyglyph'>::</span> <span class='hs-layout'>(</span><span class='hs-conid'>TensorType</span> <span class='hs-varid'>a</span><span class='hs-layout'>,</span> <span class='hs-conid'>Num</span> <span class='hs-varid'>a</span><span class='hs-layout'>)</span> <span class='hs-keyglyph'>=></span>
<a name="line-182"></a> <span class='hs-conid'>TensorFlow</span><span class='hs-varop'>.</span><span class='hs-conid'>Types</span><span class='hs-varop'>.</span><span class='hs-conid'>Shape</span> <span class='hs-keyglyph'>-></span> <span class='hs-conid'>Build</span> <span class='hs-layout'>(</span><span class='hs-conid'>Tensor</span> <span class='hs-conid'>TensorFlow</span><span class='hs-varop'>.</span><span class='hs-conid'>Tensor</span><span class='hs-varop'>.</span><span class='hs-conid'>Ref</span> <span class='hs-varid'>a</span><span class='hs-layout'>)</span>
<a name="line-183"></a><span class='hs-definition'>zeroInitializedVariable</span> <span class='hs-keyglyph'>=</span> <span class='hs-varid'>initializedVariable</span> <span class='hs-varop'>.</span> <span class='hs-varid'>zeros</span>
<a name="line-184"></a>
<a name="line-185"></a><a name="save"></a><span class='hs-comment'>-- TODO: Support heterogeneous list of tensors.</span>
<a name="line-186"></a><span class='hs-definition'>save</span> <span class='hs-keyglyph'>::</span> <span class='hs-keyword'>forall</span> <span class='hs-varid'>a</span> <span class='hs-varid'>v</span> <span class='hs-varop'>.</span> <span class='hs-conid'>TensorType</span> <span class='hs-varid'>a</span>
<a name="line-187"></a> <span class='hs-keyglyph'>=></span> <span class='hs-conid'>ByteString</span> <span class='hs-comment'>-- ^ File path.</span>
<a name="line-188"></a> <span class='hs-keyglyph'>-></span> <span class='hs-keyglyph'>[</span><span class='hs-conid'>Tensor</span> <span class='hs-varid'>v</span> <span class='hs-varid'>a</span><span class='hs-keyglyph'>]</span> <span class='hs-comment'>-- ^ Tensors to save.</span>
<a name="line-189"></a> <span class='hs-keyglyph'>-></span> <span class='hs-conid'>Build</span> <span class='hs-conid'>ControlNode</span>
<a name="line-190"></a><span class='hs-definition'>save</span> <span class='hs-varid'>path</span> <span class='hs-varid'>xs</span> <span class='hs-keyglyph'>=</span> <span class='hs-keyword'>do</span>
<a name="line-191"></a> <span class='hs-keyword'>let</span> <span class='hs-varid'>toByteStringTensor</span> <span class='hs-keyglyph'>=</span> <span class='hs-varid'>scalar</span> <span class='hs-varop'>.</span> <span class='hs-varid'>encodeUtf8</span> <span class='hs-varop'>.</span> <span class='hs-varid'>unNodeName</span>
<a name="line-192"></a> <span class='hs-varid'>names</span> <span class='hs-keyglyph'><-</span> <span class='hs-varid'>mapM</span> <span class='hs-layout'>(</span><span class='hs-varid'>fmap</span> <span class='hs-varid'>toByteStringTensor</span> <span class='hs-varop'>.</span> <span class='hs-varid'>renderNodeName</span><span class='hs-layout'>)</span> <span class='hs-varid'>xs</span>
<a name="line-193"></a> <span class='hs-keyword'>let</span> <span class='hs-varid'>types</span> <span class='hs-keyglyph'>=</span> <span class='hs-varid'>replicate</span> <span class='hs-layout'>(</span><span class='hs-varid'>length</span> <span class='hs-varid'>xs</span><span class='hs-layout'>)</span> <span class='hs-layout'>(</span><span class='hs-varid'>tensorType</span> <span class='hs-layout'>(</span><span class='hs-varid'>undefined</span> <span class='hs-keyglyph'>::</span> <span class='hs-varid'>a</span><span class='hs-layout'>)</span><span class='hs-layout'>)</span>
<a name="line-194"></a> <span class='hs-keyword'>let</span> <span class='hs-varid'>saveOp</span> <span class='hs-keyglyph'>=</span> <span class='hs-varid'>buildOp</span> <span class='hs-varop'>$</span> <span class='hs-varid'>opDef</span> <span class='hs-str'>"Save"</span>
<a name="line-195"></a> <span class='hs-varop'>&</span> <span class='hs-varid'>opAttr</span> <span class='hs-str'>"T"</span> <span class='hs-varop'>.~</span> <span class='hs-varid'>types</span>
<a name="line-196"></a> <span class='hs-varid'>saveOp</span> <span class='hs-layout'>(</span><span class='hs-varid'>scalar</span> <span class='hs-varid'>path</span><span class='hs-layout'>)</span> <span class='hs-layout'>(</span><span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>pack</span> <span class='hs-varid'>names</span><span class='hs-layout'>)</span> <span class='hs-varid'>xs</span>
<a name="line-197"></a>
<a name="line-198"></a><a name="restoreFromName"></a><span class='hs-comment'>-- | Restore a tensor's value from a checkpoint file.</span>
<a name="line-199"></a><span class='hs-comment'>--</span>
<a name="line-200"></a><span class='hs-comment'>-- This version allows restoring from a checkpoint file that uses a different</span>
<a name="line-201"></a><span class='hs-comment'>-- tensor name than the variable.</span>
<a name="line-202"></a><span class='hs-definition'>restoreFromName</span> <span class='hs-keyglyph'>::</span> <span class='hs-keyword'>forall</span> <span class='hs-varid'>a</span> <span class='hs-varop'>.</span> <span class='hs-conid'>TensorType</span> <span class='hs-varid'>a</span>
<a name="line-203"></a> <span class='hs-keyglyph'>=></span> <span class='hs-conid'>ByteString</span> <span class='hs-comment'>-- ^ File path.</span>
<a name="line-204"></a> <span class='hs-keyglyph'>-></span> <span class='hs-conid'>ByteString</span> <span class='hs-comment'>-- ^ Tensor name override.</span>
<a name="line-205"></a> <span class='hs-keyglyph'>-></span> <span class='hs-conid'>Tensor</span> <span class='hs-conid'>Ref</span> <span class='hs-varid'>a</span> <span class='hs-comment'>-- ^ Tensor to restore.</span>
<a name="line-206"></a> <span class='hs-keyglyph'>-></span> <span class='hs-conid'>Build</span> <span class='hs-conid'>ControlNode</span>
<a name="line-207"></a><span class='hs-definition'>restoreFromName</span> <span class='hs-varid'>path</span> <span class='hs-varid'>name</span> <span class='hs-varid'>x</span> <span class='hs-keyglyph'>=</span> <span class='hs-keyword'>do</span>
<a name="line-208"></a> <span class='hs-keyword'>let</span> <span class='hs-varid'>restoreOp</span> <span class='hs-keyglyph'>=</span> <span class='hs-varid'>buildOp</span> <span class='hs-varop'>$</span> <span class='hs-varid'>opDef</span> <span class='hs-str'>"Restore"</span>
<a name="line-209"></a> <span class='hs-varop'>&</span> <span class='hs-varid'>opAttr</span> <span class='hs-str'>"dt"</span> <span class='hs-varop'>.~</span> <span class='hs-varid'>tensorType</span> <span class='hs-layout'>(</span><span class='hs-varid'>undefined</span> <span class='hs-keyglyph'>::</span> <span class='hs-varid'>a</span><span class='hs-layout'>)</span>
<a name="line-210"></a> <span class='hs-varid'>group</span> <span class='hs-varop'>=<<</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>assign</span> <span class='hs-varid'>x</span>
<a name="line-211"></a> <span class='hs-layout'>(</span><span class='hs-varid'>restoreOp</span> <span class='hs-layout'>(</span><span class='hs-varid'>scalar</span> <span class='hs-varid'>path</span><span class='hs-layout'>)</span> <span class='hs-layout'>(</span><span class='hs-varid'>scalar</span> <span class='hs-varid'>name</span><span class='hs-layout'>)</span> <span class='hs-keyglyph'>::</span> <span class='hs-conid'>Tensor</span> <span class='hs-conid'>Value</span> <span class='hs-varid'>a</span><span class='hs-layout'>)</span>
<a name="line-212"></a>
<a name="line-213"></a><a name="restore"></a><span class='hs-comment'>-- | Restore a tensor's value from a checkpoint file.</span>
<a name="line-214"></a><span class='hs-definition'>restore</span> <span class='hs-keyglyph'>::</span> <span class='hs-keyword'>forall</span> <span class='hs-varid'>a</span> <span class='hs-varop'>.</span> <span class='hs-conid'>TensorType</span> <span class='hs-varid'>a</span>
<a name="line-215"></a> <span class='hs-keyglyph'>=></span> <span class='hs-conid'>ByteString</span> <span class='hs-comment'>-- ^ File path.</span>
<a name="line-216"></a> <span class='hs-keyglyph'>-></span> <span class='hs-conid'>Tensor</span> <span class='hs-conid'>Ref</span> <span class='hs-varid'>a</span> <span class='hs-comment'>-- ^ Tensor to restore.</span>
<a name="line-217"></a> <span class='hs-keyglyph'>-></span> <span class='hs-conid'>Build</span> <span class='hs-conid'>ControlNode</span>
<a name="line-218"></a><span class='hs-definition'>restore</span> <span class='hs-varid'>path</span> <span class='hs-varid'>x</span> <span class='hs-keyglyph'>=</span> <span class='hs-keyword'>do</span>
<a name="line-219"></a> <span class='hs-varid'>name</span> <span class='hs-keyglyph'><-</span> <span class='hs-varid'>encodeUtf8</span> <span class='hs-varop'>.</span> <span class='hs-varid'>unNodeName</span> <span class='hs-varop'><$></span> <span class='hs-varid'>renderNodeName</span> <span class='hs-varid'>x</span>
<a name="line-220"></a> <span class='hs-varid'>restoreFromName</span> <span class='hs-varid'>path</span> <span class='hs-varid'>name</span> <span class='hs-varid'>x</span>
<a name="line-221"></a>
<a name="line-222"></a><a name="constant"></a><span class='hs-comment'>-- | Create a constant tensor.</span>
<a name="line-223"></a><span class='hs-comment'>--</span>
<a name="line-224"></a><span class='hs-comment'>-- The values should be in row major order, e.g.,</span>
<a name="line-225"></a><span class='hs-comment'>--</span>
<a name="line-226"></a><span class='hs-comment'>-- element 0: index (0, ..., 0)</span>
<a name="line-227"></a><span class='hs-comment'>-- element 1: index (0, ..., 1)</span>
<a name="line-228"></a><span class='hs-comment'>-- ...</span>
<a name="line-229"></a><span class='hs-definition'>constant</span> <span class='hs-keyglyph'>::</span> <span class='hs-keyword'>forall</span> <span class='hs-varid'>a</span> <span class='hs-varop'>.</span> <span class='hs-conid'>TensorType</span> <span class='hs-varid'>a</span> <span class='hs-keyglyph'>=></span> <span class='hs-conid'>Shape</span> <span class='hs-keyglyph'>-></span> <span class='hs-keyglyph'>[</span><span class='hs-varid'>a</span><span class='hs-keyglyph'>]</span> <span class='hs-keyglyph'>-></span> <span class='hs-conid'>Tensor</span> <span class='hs-conid'>Value</span> <span class='hs-varid'>a</span>
<a name="line-230"></a><span class='hs-definition'>constant</span> <span class='hs-layout'>(</span><span class='hs-conid'>Shape</span> <span class='hs-varid'>shape'</span><span class='hs-layout'>)</span> <span class='hs-varid'>values</span>
<a name="line-231"></a> <span class='hs-keyglyph'>|</span> <span class='hs-varid'>invalidLength</span> <span class='hs-keyglyph'>=</span> <span class='hs-varid'>error</span> <span class='hs-varid'>invalidLengthMsg</span>
<a name="line-232"></a> <span class='hs-keyglyph'>|</span> <span class='hs-varid'>otherwise</span> <span class='hs-keyglyph'>=</span> <span class='hs-varid'>buildOp</span> <span class='hs-varop'>$</span> <span class='hs-varid'>opDef</span> <span class='hs-str'>"Const"</span>
<a name="line-233"></a> <span class='hs-varop'>&</span> <span class='hs-varid'>opAttr</span> <span class='hs-str'>"value"</span> <span class='hs-varop'>.~</span> <span class='hs-varid'>typedNode</span>
<a name="line-234"></a> <span class='hs-varop'>&</span> <span class='hs-varid'>opAttr</span> <span class='hs-str'>"dtype"</span> <span class='hs-varop'>.~</span> <span class='hs-varid'>nodeType</span>
<a name="line-235"></a> <span class='hs-keyword'>where</span>
<a name="line-236"></a> <span class='hs-varid'>invalidLength</span> <span class='hs-keyglyph'>=</span> <span class='hs-varid'>product</span> <span class='hs-varid'>shape'</span> <span class='hs-varop'>/=</span> <span class='hs-varid'>fromIntegral</span> <span class='hs-layout'>(</span><span class='hs-varid'>length</span> <span class='hs-varid'>values</span><span class='hs-layout'>)</span>
<a name="line-237"></a> <span class='hs-varid'>invalidLengthMsg</span> <span class='hs-keyglyph'>=</span> <span class='hs-varid'>printf</span> <span class='hs-str'>"invalid tensor length: expected %d got %d"</span>
<a name="line-238"></a> <span class='hs-layout'>(</span><span class='hs-varid'>product</span> <span class='hs-varid'>shape'</span><span class='hs-layout'>)</span>
<a name="line-239"></a> <span class='hs-layout'>(</span><span class='hs-varid'>length</span> <span class='hs-varid'>values</span><span class='hs-layout'>)</span>
<a name="line-240"></a> <span class='hs-varid'>nodeType</span> <span class='hs-keyglyph'>=</span> <span class='hs-varid'>tensorType</span> <span class='hs-layout'>(</span><span class='hs-varid'>undefined</span> <span class='hs-keyglyph'>::</span> <span class='hs-varid'>a</span><span class='hs-layout'>)</span>
<a name="line-241"></a> <span class='hs-varid'>typedNode</span> <span class='hs-keyglyph'>::</span> <span class='hs-conid'>TensorProto</span>
<a name="line-242"></a> <span class='hs-varid'>typedNode</span> <span class='hs-keyglyph'>=</span> <span class='hs-varid'>def</span>
<a name="line-243"></a> <span class='hs-varop'>&</span> <span class='hs-varid'>dtype</span> <span class='hs-varop'>.~</span> <span class='hs-varid'>nodeType</span>
<a name="line-244"></a> <span class='hs-varop'>&</span> <span class='hs-varid'>tensorShape</span><span class='hs-varop'>.</span><span class='hs-conid'>TensorShape</span><span class='hs-varop'>.</span><span class='hs-varid'>dim</span> <span class='hs-varop'>.~</span>
<a name="line-245"></a> <span class='hs-keyglyph'>[</span><span class='hs-varid'>def</span> <span class='hs-varop'>&</span> <span class='hs-conid'>TensorShape</span><span class='hs-varop'>.</span><span class='hs-varid'>size</span> <span class='hs-varop'>.~</span> <span class='hs-varid'>x</span> <span class='hs-keyglyph'>|</span> <span class='hs-varid'>x</span> <span class='hs-keyglyph'><-</span> <span class='hs-varid'>shape'</span><span class='hs-keyglyph'>]</span>
<a name="line-246"></a> <span class='hs-varop'>&</span> <span class='hs-varid'>tensorVal</span> <span class='hs-varop'>.~</span> <span class='hs-varid'>values</span>
<a name="line-247"></a>
<a name="line-248"></a><a name="scalarize"></a><span class='hs-comment'>-- | Reshape a N-D tensor down to a scalar.</span>
<a name="line-249"></a><span class='hs-comment'>-- </span>
<a name="line-250"></a><span class='hs-comment'>-- See `TensorFlow.GenOps.Core.reshape`.</span>
<a name="line-251"></a><span class='hs-definition'>scalarize</span> <span class='hs-keyglyph'>::</span> <span class='hs-layout'>(</span><span class='hs-conid'>TensorType</span> <span class='hs-varid'>a</span><span class='hs-layout'>)</span> <span class='hs-keyglyph'>=></span> <span class='hs-conid'>Tensor</span> <span class='hs-varid'>v</span> <span class='hs-varid'>a</span> <span class='hs-keyglyph'>-></span> <span class='hs-conid'>Tensor</span> <span class='hs-conid'>Value</span> <span class='hs-varid'>a</span>
<a name="line-252"></a><span class='hs-definition'>scalarize</span> <span class='hs-varid'>t</span> <span class='hs-keyglyph'>=</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>reshape</span> <span class='hs-varid'>t</span> <span class='hs-layout'>(</span><span class='hs-varid'>vector</span> <span class='hs-varid'>scalarShape</span><span class='hs-layout'>)</span>
<a name="line-253"></a> <span class='hs-keyword'>where</span>
<a name="line-254"></a> <span class='hs-varid'>scalarShape</span> <span class='hs-keyglyph'>=</span> <span class='hs-conid'>[]</span> <span class='hs-keyglyph'>::</span> <span class='hs-keyglyph'>[</span><span class='hs-conid'>Int32</span><span class='hs-keyglyph'>]</span>
<a name="line-255"></a>
<a name="line-256"></a>
<a name="line-257"></a><a name="vector"></a><span class='hs-comment'>-- | Create a constant vector.</span>
<a name="line-258"></a><span class='hs-definition'>vector</span> <span class='hs-keyglyph'>::</span> <span class='hs-conid'>TensorType</span> <span class='hs-varid'>a</span> <span class='hs-keyglyph'>=></span> <span class='hs-keyglyph'>[</span><span class='hs-varid'>a</span><span class='hs-keyglyph'>]</span> <span class='hs-keyglyph'>-></span> <span class='hs-conid'>Tensor</span> <span class='hs-conid'>Value</span> <span class='hs-varid'>a</span>
<a name="line-259"></a><span class='hs-definition'>vector</span> <span class='hs-varid'>xs</span> <span class='hs-keyglyph'>=</span> <span class='hs-varid'>constant</span> <span class='hs-keyglyph'>[</span><span class='hs-varid'>fromIntegral</span> <span class='hs-varop'>$</span> <span class='hs-varid'>length</span> <span class='hs-varid'>xs</span><span class='hs-keyglyph'>]</span> <span class='hs-varid'>xs</span>
<a name="line-260"></a>
<a name="line-261"></a><a name="scalar"></a><span class='hs-comment'>-- | Create a constant scalar.</span>
<a name="line-262"></a><span class='hs-definition'>scalar</span> <span class='hs-keyglyph'>::</span> <span class='hs-keyword'>forall</span> <span class='hs-varid'>a</span> <span class='hs-varop'>.</span> <span class='hs-conid'>TensorType</span> <span class='hs-varid'>a</span> <span class='hs-keyglyph'>=></span> <span class='hs-varid'>a</span> <span class='hs-keyglyph'>-></span> <span class='hs-conid'>Tensor</span> <span class='hs-conid'>Value</span> <span class='hs-varid'>a</span>
<a name="line-263"></a><span class='hs-definition'>scalar</span> <span class='hs-varid'>x</span> <span class='hs-keyglyph'>=</span> <span class='hs-varid'>constant</span> <span class='hs-conid'>[]</span> <span class='hs-keyglyph'>[</span><span class='hs-varid'>x</span><span class='hs-keyglyph'>]</span>
<a name="line-264"></a>
<a name="line-265"></a><a name="truncatedNormal"></a><span class='hs-comment'>-- Random tensor from the unit normal distribution with bounded values.</span>
<a name="line-266"></a><span class='hs-definition'>truncatedNormal</span> <span class='hs-keyglyph'>::</span> <span class='hs-keyword'>forall</span> <span class='hs-varid'>a</span> <span class='hs-varid'>v</span> <span class='hs-varop'>.</span> <span class='hs-conid'>TensorType</span> <span class='hs-varid'>a</span>
<a name="line-267"></a> <span class='hs-keyglyph'>=></span> <span class='hs-conid'>Tensor</span> <span class='hs-varid'>v</span> <span class='hs-conid'>Int64</span> <span class='hs-comment'>-- ^ Shape.</span>
<a name="line-268"></a> <span class='hs-keyglyph'>-></span> <span class='hs-conid'>Build</span> <span class='hs-layout'>(</span><span class='hs-conid'>Tensor</span> <span class='hs-conid'>Value</span> <span class='hs-varid'>a</span><span class='hs-layout'>)</span>
<a name="line-269"></a><span class='hs-definition'>truncatedNormal</span> <span class='hs-keyglyph'>=</span> <span class='hs-varid'>buildOp</span> <span class='hs-varop'>$</span> <span class='hs-varid'>opDef</span> <span class='hs-str'>"TruncatedNormal"</span>
<a name="line-270"></a> <span class='hs-varop'>&</span> <span class='hs-varid'>opAttr</span> <span class='hs-str'>"dtype"</span> <span class='hs-varop'>.~</span> <span class='hs-varid'>tensorType</span> <span class='hs-layout'>(</span><span class='hs-varid'>undefined</span> <span class='hs-keyglyph'>::</span> <span class='hs-varid'>a</span><span class='hs-layout'>)</span>
<a name="line-271"></a> <span class='hs-varop'>&</span> <span class='hs-varid'>opAttr</span> <span class='hs-str'>"T"</span> <span class='hs-varop'>.~</span> <span class='hs-varid'>tensorType</span> <span class='hs-layout'>(</span><span class='hs-varid'>undefined</span> <span class='hs-keyglyph'>::</span> <span class='hs-conid'>Int64</span><span class='hs-layout'>)</span>
<a name="line-272"></a>
<a name="line-273"></a><a name="zeros"></a><span class='hs-definition'>zeros</span> <span class='hs-keyglyph'>::</span> <span class='hs-keyword'>forall</span> <span class='hs-varid'>a</span> <span class='hs-varop'>.</span> <span class='hs-layout'>(</span><span class='hs-conid'>Num</span> <span class='hs-varid'>a</span><span class='hs-layout'>,</span> <span class='hs-conid'>TensorType</span> <span class='hs-varid'>a</span><span class='hs-layout'>)</span> <span class='hs-keyglyph'>=></span> <span class='hs-conid'>Shape</span> <span class='hs-keyglyph'>-></span> <span class='hs-conid'>Tensor</span> <span class='hs-conid'>Value</span> <span class='hs-varid'>a</span>
<a name="line-274"></a><span class='hs-definition'>zeros</span> <span class='hs-layout'>(</span><span class='hs-conid'>Shape</span> <span class='hs-varid'>shape'</span><span class='hs-layout'>)</span> <span class='hs-keyglyph'>=</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>fill</span> <span class='hs-layout'>(</span><span class='hs-varid'>vector</span> <span class='hs-varop'>$</span> <span class='hs-varid'>map</span> <span class='hs-varid'>fromIntegral</span> <span class='hs-varid'>shape'</span><span class='hs-layout'>)</span> <span class='hs-layout'>(</span><span class='hs-varid'>scalar</span> <span class='hs-num'>0</span><span class='hs-layout'>)</span>
<a name="line-275"></a>
<a name="line-276"></a><a name="shape"></a><span class='hs-definition'>shape</span> <span class='hs-keyglyph'>::</span> <span class='hs-layout'>(</span><span class='hs-conid'>TensorType</span> <span class='hs-varid'>t</span><span class='hs-layout'>)</span> <span class='hs-keyglyph'>=></span> <span class='hs-conid'>Tensor</span> <span class='hs-varid'>v1</span> <span class='hs-varid'>t</span> <span class='hs-keyglyph'>-></span> <span class='hs-conid'>Tensor</span> <span class='hs-conid'>Value</span> <span class='hs-conid'>Int32</span>
<a name="line-277"></a><span class='hs-definition'>shape</span> <span class='hs-keyglyph'>=</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>shape</span>
<a name="line-278"></a>
<a name="line-279"></a><a name="expandDims"></a><span class='hs-definition'>expandDims</span> <span class='hs-keyglyph'>::</span> <span class='hs-layout'>(</span><span class='hs-conid'>TensorType</span> <span class='hs-varid'>t</span><span class='hs-layout'>)</span> <span class='hs-keyglyph'>=></span> <span class='hs-conid'>Tensor</span> <span class='hs-varid'>v1</span> <span class='hs-varid'>t</span> <span class='hs-keyglyph'>-></span> <span class='hs-conid'>Tensor</span> <span class='hs-varid'>v2</span> <span class='hs-conid'>Int32</span> <span class='hs-keyglyph'>-></span> <span class='hs-conid'>Tensor</span> <span class='hs-conid'>Value</span> <span class='hs-varid'>t</span>
<a name="line-280"></a><span class='hs-definition'>expandDims</span> <span class='hs-keyglyph'>=</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>expandDims</span>
<a name="line-281"></a>
<a name="line-282"></a><a name="reducedShape"></a><span class='hs-comment'>-- | Helper function for reduction ops (translation of math_ops.reduced_shape).</span>
<a name="line-283"></a><span class='hs-definition'>reducedShape</span> <span class='hs-keyglyph'>::</span> <span class='hs-layout'>(</span><span class='hs-conid'>OneOf</span> <span class='hs-chr'>'</span><span class='hs-keyglyph'>[</span> <span class='hs-conid'>Int32</span><span class='hs-layout'>,</span> <span class='hs-conid'>Int64</span> <span class='hs-keyglyph'>]</span> <span class='hs-varid'>t1</span><span class='hs-layout'>,</span> <span class='hs-conid'>OneOf</span> <span class='hs-chr'>'</span><span class='hs-keyglyph'>[</span> <span class='hs-conid'>Int32</span><span class='hs-layout'>,</span> <span class='hs-conid'>Int64</span> <span class='hs-keyglyph'>]</span> <span class='hs-varid'>t2</span><span class='hs-layout'>)</span> <span class='hs-keyglyph'>=></span>
<a name="line-284"></a> <span class='hs-conid'>Tensor</span> <span class='hs-varid'>v1</span> <span class='hs-varid'>t1</span> <span class='hs-keyglyph'>-></span> <span class='hs-conid'>Tensor</span> <span class='hs-varid'>v2</span> <span class='hs-varid'>t2</span> <span class='hs-keyglyph'>-></span> <span class='hs-conid'>Tensor</span> <span class='hs-conid'>Value</span> <span class='hs-conid'>Int32</span>
<a name="line-285"></a><span class='hs-definition'>reducedShape</span> <span class='hs-varid'>inputShape</span> <span class='hs-varid'>axes</span> <span class='hs-keyglyph'>=</span>
<a name="line-286"></a> <span class='hs-keyword'>let</span> <span class='hs-varid'>inputShape32</span> <span class='hs-keyglyph'>=</span> <span class='hs-varid'>toInt32</span> <span class='hs-varid'>inputShape</span> <span class='hs-comment'>-- [2, 3, 5, 7]</span>
<a name="line-287"></a> <span class='hs-varid'>axes32</span> <span class='hs-keyglyph'>=</span> <span class='hs-varid'>toInt32</span> <span class='hs-varid'>axes</span> <span class='hs-comment'>-- [1, 2]</span>
<a name="line-288"></a> <span class='hs-varid'>toInt32</span> <span class='hs-varid'>x</span> <span class='hs-keyglyph'>=</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>cast</span> <span class='hs-varid'>x</span> <span class='hs-keyglyph'>::</span> <span class='hs-conid'>Tensor</span> <span class='hs-conid'>Value</span> <span class='hs-conid'>Int32</span>
<a name="line-289"></a> <span class='hs-varid'>inputRank</span> <span class='hs-keyglyph'>=</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>size</span> <span class='hs-varid'>inputShape32</span> <span class='hs-comment'>-- 4</span>
<a name="line-290"></a> <span class='hs-varid'>axesMod</span> <span class='hs-keyglyph'>=</span> <span class='hs-layout'>(</span><span class='hs-varid'>axes32</span> <span class='hs-varop'>+</span> <span class='hs-varid'>inputRank</span><span class='hs-layout'>)</span> <span class='hs-varop'>`</span><span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>mod</span><span class='hs-varop'>`</span> <span class='hs-varid'>inputRank</span>
<a name="line-291"></a> <span class='hs-varid'>axesShape</span> <span class='hs-keyglyph'>=</span> <span class='hs-varid'>shape</span> <span class='hs-varid'>axesMod</span> <span class='hs-comment'>-- [2]</span>
<a name="line-292"></a> <span class='hs-keyword'>in</span> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>dynamicStitch</span> <span class='hs-comment'>-- [2, 1, 1, 7]</span>
<a name="line-293"></a> <span class='hs-keyglyph'>[</span><span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>range</span> <span class='hs-num'>0</span> <span class='hs-varid'>inputRank</span> <span class='hs-num'>1</span><span class='hs-layout'>,</span> <span class='hs-comment'>-- [0, 1, 2, 3]</span>
<a name="line-294"></a> <span class='hs-varid'>axesMod</span><span class='hs-keyglyph'>]</span> <span class='hs-comment'>-- [1, 2]</span>
<a name="line-295"></a> <span class='hs-keyglyph'>[</span><span class='hs-varid'>inputShape32</span><span class='hs-layout'>,</span> <span class='hs-comment'>-- [2, 3, 5, 7]</span>
<a name="line-296"></a> <span class='hs-conid'>CoreOps</span><span class='hs-varop'>.</span><span class='hs-varid'>fill</span> <span class='hs-varid'>axesShape</span> <span class='hs-num'>1</span><span class='hs-keyglyph'>]</span> <span class='hs-comment'>-- [1, 1]</span>
</pre></body>
</html>