From c284e02f1f03626071e2d6ba2a7cf449cb1863f3 Mon Sep 17 00:00:00 2001
From: auphelia <jakobapk@web.de>
Date: Tue, 4 Feb 2020 12:35:28 +0000
Subject: [PATCH] [Notebook] Added additional information to streamlining
section
---
notebooks/9-FINN-EndToEndFlow.ipynb | 71 ++++++++++++++++++++++++++++-
1 file changed, 70 insertions(+), 1 deletion(-)
diff --git a/notebooks/9-FINN-EndToEndFlow.ipynb b/notebooks/9-FINN-EndToEndFlow.ipynb
index 6d1360aac..9e5761b83 100644
--- a/notebooks/9-FINN-EndToEndFlow.ipynb
+++ b/notebooks/9-FINN-EndToEndFlow.ipynb
@@ -354,7 +354,76 @@
"cell_type": "markdown",
"metadata": {},
"source": [
- "As can be seen, several transformations are involved in the streamlining transformation. There are move and collapse transformations. In the last step the operations are transformed into multithresholds. The involved transformations can be viewed in detail [here](https://github.com/Xilinx/finn/tree/dev/src/finn/transformation/streamline). After each transformation, three of the basic transformations (`GiveUniqueNodeNames`, `GiveReadableTensorNames` and `InferDataTypes`) are applied to the model as clean up."
+ "As can be seen, several transformations are involved in the streamlining transformation. There are move and collapse transformations. In the last step the operations are transformed into multithresholds. The involved transformations can be viewed in detail [here](https://github.com/Xilinx/finn/tree/dev/src/finn/transformation/streamline). After each transformation, three of the basic transformations (`GiveUniqueNodeNames`, `GiveReadableTensorNames` and `InferDataTypes`) are applied to the model as clean up.\n",
+ "\n",
+ "After streamlining the network looks as follows."
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 9,
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "\n",
+ "Stopping http://0.0.0.0:8081\n",
+ "Serving 'lfc_w1_a1.onnx' at http://0.0.0.0:8081\n"
+ ]
+ }
+ ],
+ "source": [
+ "model = model.transform(Streamline())\n",
+ "model.save(\"lfc_w1_a1.onnx\")\n",
+ "netron.start(\"lfc_w1_a1.onnx\", port=8081, host=\"0.0.0.0\")"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 10,
+ "metadata": {},
+ "outputs": [
+ {
+ "data": {
+ "text/html": [
+ "<iframe src=\"http://0.0.0.0:8081/\" style=\"position: relative; width: 100%;\" height=\"400\"></iframe>\n"
+ ],
+ "text/plain": [
+ "<IPython.core.display.HTML object>"
+ ]
+ },
+ "metadata": {},
+ "output_type": "display_data"
+ }
+ ],
+ "source": [
+ "%%html\n",
+ "<iframe src=\"http://0.0.0.0:8081/\" style=\"position: relative; width: 100%;\" height=\"400\"></iframe>"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "Our example network is a quantized network with 1 bit precision. For this reason, after streamlining, the resulting bipolar matrix multiplications are converted into xnorpopcount operations. This transformation produces operations that are again collapsed and converted into thresholds. This procedure is shown below. After these transformations, the nodes can be converted to HLS layers for further processing."
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 11,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "from finn.transformation.bipolar_to_xnor import ConvertBipolarMatMulToXnorPopcount\n",
+ "import finn.transformation.streamline.absorb as absorb\n",
+ "from finn.transformation.streamline.round_thresholds import RoundAndClipThresholds\n",
+ "\n",
+ "model = model.transform(ConvertBipolarMatMulToXnorPopcount())\n",
+ "model = model.transform(absorb.AbsorbAddIntoMultiThreshold())\n",
+ "model = model.transform(absorb.AbsorbMulIntoMultiThreshold())\n",
+ "model = model.transform(RoundAndClipThresholds())"
]
},
{
--
GitLab