docs(//core/lowering): Document new aten::addmm fusion pass

Signed-off-by: Naren Dasan <naren@narendasan.com>
Signed-off-by: Naren Dasan <narens@nvidia.com>

Author: narendasan

docs/_cpp_api/structtrtorch_1_1ExtraInfo.html

@@ -960,10 +960,7 @@ <h2 id="struct-documentation">
             <a class="reference internal" href="classtrtorch_1_1ExtraInfo_1_1DataType.html#_CPPv4N7trtorch9ExtraInfo8DataTypeE" title="trtorch::ExtraInfo::DataType">
              DataType
             </a>
-            ::
-            <a class="reference internal" href="classtrtorch_1_1ExtraInfo_1_1DataType.html#_CPPv4N7trtorch9ExtraInfo8DataType5Value6kFloatE" title="trtorch::ExtraInfo::DataType::kFloat">
-             kFloat
-            </a>
+            ::kFloat
             <a class="headerlink" href="#_CPPv4N7trtorch9ExtraInfo12op_precisionE" title="Permalink to this definition">
              ¶
             </a>
@@ -1099,10 +1096,7 @@ <h2 id="struct-documentation">
             <a class="reference internal" href="classtrtorch_1_1ExtraInfo_1_1DeviceType.html#_CPPv4N7trtorch9ExtraInfo10DeviceTypeE" title="trtorch::ExtraInfo::DeviceType">
              DeviceType
             </a>
-            ::
-            <a class="reference internal" href="classtrtorch_1_1ExtraInfo_1_1DeviceType.html#_CPPv4N7trtorch9ExtraInfo10DeviceType5Value4kGPUE" title="trtorch::ExtraInfo::DeviceType::kGPU">
-             kGPU
-            </a>
+            ::kGPU
             <a class="headerlink" href="#_CPPv4N7trtorch9ExtraInfo6deviceE" title="Permalink to this definition">
              ¶
             </a>

docs/_sources/contributors/lowering.rst.txt

@@ -56,6 +56,31 @@ Freeze Module
 
 Freeze attributes and inline constants and modules. Propogates constants in the graph.
 
+Fuse AddMM Branches
+***************************************
+
+    `trtorch/core/lowering/passes/fuse_addmm_branches.cpp <https://github.com/nvidia/trtorch/blob/master/core/lowering/passes/fuse_addmm_branches.cpp>`_
+
+A common pattern in scripted modules is tensors of different dimensions use different constructions for implementing linear layers. We fuse these
+different varients into a single one that will get caught by the Unpack AddMM pass.
+
+.. code-block:: none
+
+    %ret : Tensor = prim::If(%622)
+    block0():
+      %ret.1 : Tensor = aten::addmm(%self.fc.bias, %x9.1, %3677, %3, %3)
+      -> (%ret.1)
+    block1():
+      %output.1 : Tensor = aten::matmul(%x9.1, %3677)
+      %output0.1 : Tensor = aten::add_(%output.1, %self.fc.bias, %3)
+      -> (%output0.1)
+
+We fuse this set of blocks into a graph like this:
+
+.. code-block:: none
+
+    %ret : Tensor = aten::addmm(%self.fc.bias, %x9.1, %3677, %3, %3)
+
 Fuse Linear
 ***************************************
 

docs/contributors/lowering.html

@@ -512,6 +512,11 @@
                   Freeze Module
                  </a>
                 </li>
+                <li class="md-nav__item">
+                 <a class="md-nav__link" href="#fuse-addmm-branches">
+                  Fuse AddMM Branches
+                 </a>
+                </li>
                 <li class="md-nav__item">
                  <a class="md-nav__link" href="#fuse-linear">
                   Fuse Linear
@@ -690,6 +695,47 @@ <h3 id="freeze-module">
        <p>
         Freeze attributes and inline constants and modules. Propogates constants in the graph.
        </p>
+       <h3 id="fuse-addmm-branches">
+        Fuse AddMM Branches
+        <a class="headerlink" href="#fuse-addmm-branches" title="Permalink to this headline">
+         ¶
+        </a>
+       </h3>
+       <blockquote>
+        <div>
+         <p>
+          <a class="reference external" href="https://github.com/nvidia/trtorch/blob/master/core/lowering/passes/fuse_addmm_branches.cpp">
+           trtorch/core/lowering/passes/fuse_addmm_branches.cpp
+          </a>
+         </p>
+        </div>
+       </blockquote>
+       <p>
+        A common pattern in scripted modules is tensors of different dimensions use different constructions for implementing linear layers. We fuse these
+different varients into a single one that will get caught by the Unpack AddMM pass.
+       </p>
+       <div class="highlight-none notranslate">
+        <div class="highlight">
+         <pre><span></span>%ret : Tensor = prim::If(%622)
+block0():
+  %ret.1 : Tensor = aten::addmm(%self.fc.bias, %x9.1, %3677, %3, %3)
+  -&gt; (%ret.1)
+block1():
+  %output.1 : Tensor = aten::matmul(%x9.1, %3677)
+  %output0.1 : Tensor = aten::add_(%output.1, %self.fc.bias, %3)
+  -&gt; (%output0.1)
+</pre>
+        </div>
+       </div>
+       <p>
+        We fuse this set of blocks into a graph like this:
+       </p>
+       <div class="highlight-none notranslate">
+        <div class="highlight">
+         <pre><span></span>%ret : Tensor = aten::addmm(%self.fc.bias, %x9.1, %3677, %3, %3)
+</pre>
+        </div>
+       </div>
        <h3 id="fuse-linear">
         Fuse Linear
         <a class="headerlink" href="#fuse-linear" title="Permalink to this headline">

docs/objects.inv

@@ -1 +1 @@
-<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9"><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/class_view_hierarchy.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/classtrtorch_1_1ExtraInfo_1_1DataType.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/classtrtorch_1_1ExtraInfo_1_1DeviceType.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/classtrtorch_1_1ptq_1_1Int8CacheCalibrator.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/classtrtorch_1_1ptq_1_1Int8Calibrator.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/define_macros_8h_1a18d295a837ac71add5578860b55e5502.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/define_macros_8h_1a20c1fbeb21757871c52299dc52351b5f.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/define_macros_8h_1a25ee153c325dfc7466a33cbd5c1ff055.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/define_macros_8h_1a48d6029a45583a06848891cb0e86f7ba.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/define_macros_8h_1a71b02dddfabe869498ad5a88e11c440f.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/define_macros_8h_1a9d31d0569348d109b1b069b972dd143e.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/define_macros_8h_1abe87b341f562fd1cf40b7672e4d759da.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/define_macros_8h_1ae1c56ab8a40af292a9a4964651524d84.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/dir_cpp.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/dir_cpp_api.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/dir_cpp_api_include.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/dir_cpp_api_include_trtorch.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/enum_logging_8h_1a5f612ff2f783ff4fbe89d168f0d817d4.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/file_cpp_api_include_trtorch_logging.h.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/file_cpp_api_include_trtorch_macros.h.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/file_cpp_api_include_trtorch_ptq.h.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/file_cpp_api_include_trtorch_trtorch.h.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/file_view_hierarchy.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/function_logging_8h_1a118d65b179defff7fff279eb9cd126cb.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/function_logging_8h_1a396a688110397538f8b3fb7dfdaf38bb.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/function_logging_8h_1a9b420280bfacc016d7e36a5704021949.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/function_logging_8h_1aa533955a2b908db9e5df5acdfa24715f.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/function_logging_8h_1abc57d473f3af292551dee8b9c78373ad.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/function_logging_8h_1adf5435f0dbb09c0d931a1b851847236b.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/function_logging_8h_1aef44b69c62af7cf2edc8875a9506641a.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/function_trtorch_8h_1a2cf17d43ba9117b3b4d652744b4f0447.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/function_trtorch_8h_1a4422781719d7befedb364cacd91c6247.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/function_trtorch_8h_1a536bba54b70e44554099d23fa3d7e804.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/function_trtorch_8h_1a5f33b142bc2f3f2aaf462270b3ad7e31.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/function_trtorch_8h_1a726f6e7091b6b7be45b5a4275b2ffb10.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/function_trtorch_8h_1ab01696cfe08b6a5293c55935a9713c25.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/function_trtorch_8h_1ae38897d1ca4438227c970029d0f76fb5.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/namespace_trtorch.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/namespace_trtorch__logging.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/namespace_trtorch__ptq.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/program_listing_file_cpp_api_include_trtorch_logging.h.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/program_listing_file_cpp_api_include_trtorch_macros.h.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/program_listing_file_cpp_api_include_trtorch_ptq.h.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/program_listing_file_cpp_api_include_trtorch_trtorch.h.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/structtrtorch_1_1ExtraInfo.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/structtrtorch_1_1ExtraInfo_1_1InputRange.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/trtorch_cpp.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/unabridged_api.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/unabridged_orphan.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/contributors/conversion.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/contributors/execution.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/contributors/lowering.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/contributors/phases.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/contributors/system_overview.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/contributors/useful_links.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/contributors/writing_converters.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/index.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/py_api/logging.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/py_api/trtorch.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/tutorials/getting_started.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/tutorials/installation.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/tutorials/ptq.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/tutorials/trtorchc.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/genindex.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/py-modindex.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/search.html</loc></url></urlset>
+<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9"><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/class_view_hierarchy.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/classtrtorch_1_1ExtraInfo_1_1DataType.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/classtrtorch_1_1ExtraInfo_1_1DeviceType.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/classtrtorch_1_1ptq_1_1Int8CacheCalibrator.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/classtrtorch_1_1ptq_1_1Int8Calibrator.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/define_macros_8h_1a18d295a837ac71add5578860b55e5502.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/define_macros_8h_1a20c1fbeb21757871c52299dc52351b5f.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/define_macros_8h_1a25ee153c325dfc7466a33cbd5c1ff055.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/define_macros_8h_1a48d6029a45583a06848891cb0e86f7ba.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/define_macros_8h_1a71b02dddfabe869498ad5a88e11c440f.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/define_macros_8h_1a9d31d0569348d109b1b069b972dd143e.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/define_macros_8h_1abe87b341f562fd1cf40b7672e4d759da.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/define_macros_8h_1ae1c56ab8a40af292a9a4964651524d84.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/dir_cpp.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/dir_cpp_api.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/dir_cpp_api_include.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/dir_cpp_api_include_trtorch.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/enum_logging_8h_1a5f612ff2f783ff4fbe89d168f0d817d4.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/file_cpp_api_include_trtorch_logging.h.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/file_cpp_api_include_trtorch_macros.h.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/file_cpp_api_include_trtorch_ptq.h.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/file_cpp_api_include_trtorch_trtorch.h.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/file_view_hierarchy.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/function_logging_8h_1a118d65b179defff7fff279eb9cd126cb.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/function_logging_8h_1a396a688110397538f8b3fb7dfdaf38bb.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/function_logging_8h_1a9b420280bfacc016d7e36a5704021949.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/function_logging_8h_1aa533955a2b908db9e5df5acdfa24715f.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/function_logging_8h_1abc57d473f3af292551dee8b9c78373ad.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/function_logging_8h_1adf5435f0dbb09c0d931a1b851847236b.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/function_logging_8h_1aef44b69c62af7cf2edc8875a9506641a.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/function_trtorch_8h_1a2cf17d43ba9117b3b4d652744b4f0447.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/function_trtorch_8h_1a4422781719d7befedb364cacd91c6247.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/function_trtorch_8h_1a536bba54b70e44554099d23fa3d7e804.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/function_trtorch_8h_1a5f33b142bc2f3f2aaf462270b3ad7e31.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/function_trtorch_8h_1a726f6e7091b6b7be45b5a4275b2ffb10.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/function_trtorch_8h_1ab01696cfe08b6a5293c55935a9713c25.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/function_trtorch_8h_1ae38897d1ca4438227c970029d0f76fb5.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/namespace_trtorch.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/namespace_trtorch__logging.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/namespace_trtorch__ptq.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/program_listing_file_cpp_api_include_trtorch_logging.h.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/program_listing_file_cpp_api_include_trtorch_macros.h.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/program_listing_file_cpp_api_include_trtorch_ptq.h.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/program_listing_file_cpp_api_include_trtorch_trtorch.h.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/structtrtorch_1_1ExtraInfo.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/structtrtorch_1_1ExtraInfo_1_1InputRange.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/trtorch_cpp.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/unabridged_api.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/_cpp_api/unabridged_orphan.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/contributors/lowering.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/contributors/phases.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/contributors/system_overview.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/index.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/py_api/trtorch.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/tutorials/getting_started.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/tutorials/trtorchc.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/genindex.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/py-modindex.html</loc></url><url><loc>https://nvidia.github.io/TRTorch/search.html</loc></url></urlset>

docs/searchindex.js

@@ -56,6 +56,31 @@ Freeze Module
 
 Freeze attributes and inline constants and modules. Propogates constants in the graph.
 
+Fuse AddMM Branches
+***************************************
+
+    `trtorch/core/lowering/passes/fuse_addmm_branches.cpp <https://github.com/nvidia/trtorch/blob/master/core/lowering/passes/fuse_addmm_branches.cpp>`_
+
+A common pattern in scripted modules is tensors of different dimensions use different constructions for implementing linear layers. We fuse these
+different varients into a single one that will get caught by the Unpack AddMM pass.
+
+.. code-block:: none
+
+    %ret : Tensor = prim::If(%622)
+    block0():
+      %ret.1 : Tensor = aten::addmm(%self.fc.bias, %x9.1, %3677, %3, %3)
+      -> (%ret.1)
+    block1():
+      %output.1 : Tensor = aten::matmul(%x9.1, %3677)
+      %output0.1 : Tensor = aten::add_(%output.1, %self.fc.bias, %3)
+      -> (%output0.1)
+
+We fuse this set of blocks into a graph like this:
+
+.. code-block:: none
+
+    %ret : Tensor = aten::addmm(%self.fc.bias, %x9.1, %3677, %3, %3)
+
 Fuse Linear
 ***************************************