Fixed examples

Author: stephenchouca

docs/data_analytics/index.html

@@ -71,6 +71,21 @@
 
             <li class="toctree-l1">
 
+    <span class="caption-text">Python Library</span>
+    <ul class="subnav">
+                <li class="">
+                    
+    <a class="" href="../pytensors/">Defining Tensors</a>
+                </li>
+                <li class="">
+                    
+    <a class="" href="../pycomputations/">Computing on Tensors</a>
+                </li>
+    </ul>
+	    </li>
+          
+            <li class="toctree-l1">
+		
     <span class="caption-text">Example Applications</span>
     <ul class="subnav">
                 <li class="">
@@ -133,7 +148,7 @@
 
 <p>You can use the taco C++ library to easily and efficiently compute the MTTKRP as demonstrated here:</p>
 <pre><code class="c++">// On Linux and MacOS, you can compile and run this program like so:
-//   g++ -std=c++11 -O3 -DNDEBUG -DTACO -I ../../include -L../../build/lib -ltaco mttkrp.cpp -o mttkrp
+//   g++ -std=c++11 -O3 -DNDEBUG -DTACO -I ../../include -L../../build/lib mttkrp.cpp -o mttkrp -ltaco
 //   LD_LIBRARY_PATH=../../build/lib ./mttkrp
 
 #include &lt;random&gt;
@@ -203,6 +218,34 @@
 }
 </code></pre>
 
+<p>We can also express this using the Python API.</p>
+<pre><code class="python">import pytaco as pt
+import numpy as np
+from pytaco import compressed, dense, format
+
+# Declare tensor formats
+csf = format([compressed, compressed, compressed])
+rm  = format([dense, dense])
+
+# Load a sparse order-3 tensor from file (stored in the FROSTT format) and 
+# store it as a compressed sparse fiber tensor. The tensor in this example 
+# can be download from: http://frostt.io/tensors/nell-2/
+B = pt.read(&quot;nell-2.tns&quot;, csf);
+
+# Use numpy to create random matrices
+C = pt.from_numpy_array(np.random.uniform( size=(B.shape[1], 25) ) )
+D = pt.from_numpy_array(np.random.uniform( size=(B.shape[2], 25) ) )
+
+# Create output tensor
+A = pt.tensor([B.shape[0], 25], rm)
+
+# Create index vars and define the MTTKRP op
+i, j, k, l = get_index_vars(4)
+A[i, j] = B[i, k, l] * D[l, j] * C[k, j]
+
+pt.write(&quot;A.tns&quot;, A)
+</code></pre>
+
 <p>Under the hood, when you run the above C++ program, taco generates the imperative code shown below to compute the MTTKRP. taco is able to evaluate this compound operation efficiently with a single kernel that avoids materializing the intermediate Khatri-Rao product.</p>
 <pre><code class="c++">for (int B1_pos = B.d1.pos[0]; B1_pos &lt; B.d1.pos[(0 + 1)]; B1_pos++) {
   int iB = B.d1.idx[B1_pos];

docs/machine_learning/index.html

@@ -71,6 +71,21 @@
 
             <li class="toctree-l1">
 
+    <span class="caption-text">Python Library</span>
+    <ul class="subnav">
+                <li class="">
+                    
+    <a class="" href="../pytensors/">Defining Tensors</a>
+                </li>
+                <li class="">
+                    
+    <a class="" href="../pycomputations/">Computing on Tensors</a>
+                </li>
+    </ul>
+	    </li>
+          
+            <li class="toctree-l1">
+		
     <span class="caption-text">Example Applications</span>
     <ul class="subnav">
                 <li class="">
@@ -134,7 +149,7 @@
 
 <p>You can use the taco C++ library to easily and efficiently compute the SDDMM as demonstrated here:</p>
 <pre><code class="c++">// On Linux and MacOS, you can compile and run this program like so:
-//   g++ -std=c++11 -O3 -DNDEBUG -DTACO -I ../../include -L../../build/lib -ltaco sddmm.cpp -o sddmm
+//   g++ -std=c++11 -O3 -DNDEBUG -DTACO -I ../../include -L../../build/lib sddmm.cpp -o sddmm -ltaco
 //   LD_LIBRARY_PATH=../../build/lib ./sddmm
 
 #include &lt;random&gt;
@@ -204,6 +219,40 @@
 }
 </code></pre>
 
+<p>We can also express this using the Python API:</p>
+<pre><code class="python">import pytaco as pt
+from pytaco import dense, compressed, format
+import numpy as np
+
+# Predeclare the storage formats that the inputs and output will be stored as.
+# To define a format, you must specify whether each dimension is dense or sparse 
+# and (optionally) the order in which dimensions should be stored. The formats 
+# declared below correspond to doubly compressed sparse row (dcsr), row-major 
+# dense (rm), and column-major dense (dm).
+dcsr = format([compressed, compressed])
+rm   = format([dense, dense])
+cm   = format([dense, dense], [1, 0])
+
+
+# The matrix in this example can be download from:
+# https://www.cise.ufl.edu/research/sparse/MM/Williams/webbase-1M.tar.gz
+B = pt.read(&quot;webbase-1M.mtx&quot;, dcsr)
+
+# Use numpy to create random matrices
+x = pt.from_numpy_array(np.random.uniform( size=(B.shape[0], 1000) ) )
+z = pt.from_numpy_array(np.random.uniform( size=(1000, B.shape[1]) ), out_format=cm )
+
+# Declare output matrix as doubly compressed sparse row
+A = pt.tensor(B.shape, dcsr)
+
+# Create index vars
+i, j, k = pt.get_index_vars(3)
+A[i, j] = B[i, j] * C[i, k] * D[k, j]
+
+# store tensor
+pt.write(&quot;A.mtx&quot;, A)
+</code></pre>
+
 <p>Under the hood, when you run the above C++ program, taco generates the imperative code shown below to compute the SDDMM. taco is able to do this efficiently by only computing entries of the intermediate matrix product that are actually needed to compute the output tensor <code>A</code>.</p>
 <pre><code class="c++">int A1_pos = A.d1.pos[0];
 int A2_pos = A.d2.pos[A1_pos];

docs/scientific_computing/index.html

@@ -71,6 +71,21 @@
 
             <li class="toctree-l1">
 
+    <span class="caption-text">Python Library</span>
+    <ul class="subnav">
+                <li class="">
+                    
+    <a class="" href="../pytensors/">Defining Tensors</a>
+                </li>
+                <li class="">
+                    
+    <a class="" href="../pycomputations/">Computing on Tensors</a>
+                </li>
+    </ul>
+	    </li>
+          
+            <li class="toctree-l1">
+		
     <span class="caption-text">Example Applications</span>
     <ul class="subnav">
                 <li class=" current">
@@ -133,7 +148,7 @@
 
 <p>You can use the taco C++ library to easily and efficiently compute the SpMV as demonstrated here:</p>
 <pre><code class="c++">// On Linux and MacOS, you can compile and run this program like so:
-//   g++ -std=c++11 -O3 -DNDEBUG -DTACO -I ../../include -L../../build/lib -ltaco spmv.cpp -o spmv
+//   g++ -std=c++11 -O3 -DNDEBUG -DTACO -I ../../include -L../../build/lib spmv.cpp -o spmv -ltaco
 //   LD_LIBRARY_PATH=../../build/lib ./spmv
 
 #include &lt;random&gt;
@@ -162,14 +177,14 @@
   // Generate a random dense vector and store it in the dense vector format. 
   // Vectors correspond to order-1 tensors in taco.
   Tensor&lt;double&gt; x({A.getDimension(1)}, dv);
-  for (int i = 0; i &lt; x.getDimension(0)]; ++i) {
+  for (int i = 0; i &lt; x.getDimension(0); ++i) {
     x.insert({i}, unif(gen));
   }
   x.pack();
 
   // Generate another random dense vetor and store it in the dense vector format..
   Tensor&lt;double&gt; z({A.getDimension(0)}, dv);
-  for (int i = 0; i &lt; z.getDimension(0)]; ++i) {
+  for (int i = 0; i &lt; z.getDimension(0); ++i) {
     z.insert({i}, unif(gen));
   }
   z.pack();
@@ -203,6 +218,37 @@
 }
 </code></pre>
 
+<p>You can also use the Python library to compute SpMV as shown here:</p>
+<pre><code class="python">import pytaco as pt
+from pytaco import compressed, dense
+import numpy as np
+
+# Declare the storage formats as explained in the C++ sample
+csr = pt.format([dense, compressed])
+dv  = pt.format([dense])
+
+# Load a sparse matrix stored in the matrix market format) and store it as a csr matrix. 
+# The matrix  # in this example can be downloaded from:
+# https://www.cise.ufl.edu/research/sparse/MM/Boeing/pwtk.tar.gz
+A = pt.read(&quot;pwtk.mtx&quot;, csr)
+
+# Generate two random vectors using numpy and pass them into taco
+x = pt.from_numpy_array(np.random.uniform(size=A.shape[0]))
+z = pt.from_numpy_array(np.random.uniform(size=A.shape[0]))
+
+# Declare output vector as dense
+y = pt.tensor([A.shape[0]], dv)
+
+# Create index vars
+i, j = pt.get_index_vars(2)
+
+# Define the SpMV computation
+y[i] = 42 * A[i, j] * x[j] + 33 * z[i]
+
+# Store the output
+pt.write(&quot;y.tns&quot;, y)
+</code></pre>
+
 <p>Under the hood, when you run the above C++ program, taco generates the imperative code shown below to compute the SpMV. taco is able to evaluate this compound operation efficiently with a single kernel that avoids materializing the intermediate matrix-vector product.</p>
 <pre><code class="c++">for (int iA = 0; iA &lt; 217918; iA++) {
   double tj = 0;
@@ -225,7 +271,7 @@
         <a href="../machine_learning/" class="btn btn-neutral float-right" title="Machine Learning: SDDMM">Next <span class="icon icon-circle-arrow-right"></span></a>
 
 
-        <a href="../computations/" class="btn btn-neutral" title="Computing on Tensors"><span class="icon icon-circle-arrow-left"></span> Previous</a>
+        <a href="../pycomputations/" class="btn btn-neutral" title="Computing on Tensors"><span class="icon icon-circle-arrow-left"></span> Previous</a>
 
     </div>
 
@@ -259,7 +305,7 @@
     <span class="rst-current-version" data-toggle="rst-current-version">
 
 
-        <span><a href="../computations/" style="color: #fcfcfc;">&laquo; Previous</a></span>
+        <span><a href="../pycomputations/" style="color: #fcfcfc;">&laquo; Previous</a></span>
 
 
         <span style="margin-left: 15px"><a href="../machine_learning/" style="color: #fcfcfc">Next &raquo;</a></span>

documentation/docs/data_analytics.md

@@ -8,7 +8,7 @@ You can use the taco C++ library to easily and efficiently compute the MTTKRP as
 
 ```c++
 // On Linux and MacOS, you can compile and run this program like so:
-//   g++ -std=c++11 -O3 -DNDEBUG -DTACO -I ../../include -L../../build/lib -ltaco mttkrp.cpp -o mttkrp
+//   g++ -std=c++11 -O3 -DNDEBUG -DTACO -I ../../include -L../../build/lib mttkrp.cpp -o mttkrp -ltaco
 //   LD_LIBRARY_PATH=../../build/lib ./mttkrp
 
 #include <random>

documentation/docs/machine_learning.md

@@ -9,7 +9,7 @@ You can use the taco C++ library to easily and efficiently compute the SDDMM as
 
 ```c++
 // On Linux and MacOS, you can compile and run this program like so:
-//   g++ -std=c++11 -O3 -DNDEBUG -DTACO -I ../../include -L../../build/lib -ltaco sddmm.cpp -o sddmm
+//   g++ -std=c++11 -O3 -DNDEBUG -DTACO -I ../../include -L../../build/lib sddmm.cpp -o sddmm -ltaco
 //   LD_LIBRARY_PATH=../../build/lib ./sddmm
 
 #include <random>

documentation/docs/scientific_computing.md

@@ -8,7 +8,7 @@ You can use the taco C++ library to easily and efficiently compute the SpMV as d
 
 ```c++
 // On Linux and MacOS, you can compile and run this program like so:
-//   g++ -std=c++11 -O3 -DNDEBUG -DTACO -I ../../include -L../../build/lib -ltaco spmv.cpp -o spmv
+//   g++ -std=c++11 -O3 -DNDEBUG -DTACO -I ../../include -L../../build/lib spmv.cpp -o spmv -ltaco
 //   LD_LIBRARY_PATH=../../build/lib ./spmv
 
 #include <random>
@@ -37,14 +37,14 @@ int main(int argc, char* argv[]) {
   // Generate a random dense vector and store it in the dense vector format. 
   // Vectors correspond to order-1 tensors in taco.
   Tensor<double> x({A.getDimension(1)}, dv);
-  for (int i = 0; i < x.getDimension(0)]; ++i) {
+  for (int i = 0; i < x.getDimension(0); ++i) {
     x.insert({i}, unif(gen));
   }
   x.pack();
 
   // Generate another random dense vetor and store it in the dense vector format..
   Tensor<double> z({A.getDimension(0)}, dv);
-  for (int i = 0; i < z.getDimension(0)]; ++i) {
+  for (int i = 0; i < z.getDimension(0); ++i) {
     z.insert({i}, unif(gen));
   }
   z.pack();