From c1233860b952a44f71fdecaa4d343362f15440db Mon Sep 17 00:00:00 2001
From: auphelia <jakobapk@web.de>
Date: Tue, 7 Feb 2023 18:16:21 +0000
Subject: [PATCH] [Notebooks] Change board execution section in tfc notebook
---
.../bnn-pynq/tfc_end2end_example.ipynb | 191 +++++-------------
1 file changed, 54 insertions(+), 137 deletions(-)
diff --git a/notebooks/end2end_example/bnn-pynq/tfc_end2end_example.ipynb b/notebooks/end2end_example/bnn-pynq/tfc_end2end_example.ipynb
index 851b8998d..e6fbc7f13 100644
--- a/notebooks/end2end_example/bnn-pynq/tfc_end2end_example.ipynb
+++ b/notebooks/end2end_example/bnn-pynq/tfc_end2end_example.ipynb
@@ -798,23 +798,21 @@
"source": [
"## 4. PYNQ deployment <a id='hw_test'></a>\n",
"\n",
- "* [Deployment and Remote Execution](#deploy)\n",
+ "* [Deployment](#deploy)\n",
"* [Validation on PYNQ Board](#validation)\n",
"* [Throughput Test on PYNQ Board](#throughput)\n",
"\n",
"\n",
- "We are almost done preparing our hardware design. We'll now put it in a form suitable for use as a PYNQ overlay, synthesize and deploy it."
+ "The bitfile and generated driver will be copied together with some necessary files for execution into a deployment folder which then can be used to run the network on the PYNQ board."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
- "### Deployment and Remote Execution <a id='deploy'></a>\n",
+ "### Deployment <a id='deploy'></a>\n",
"\n",
- "We'll now use the `DeployToPYNQ` transformation to create a deployment folder with the bitfile and driver file(s), and copy that to the PYNQ board. You can change the default IP address, username, password and target folder for the PYNQ below.\n",
- "\n",
- "**Make sure you've [set up the SSH keys for your PYNQ board](https://finn-dev.readthedocs.io/en/latest/getting_started.html#pynq-board-first-time-setup) before executing this step.**"
+ "We'll now create a deployment folder with the bitfile and driver file(s), we zip it and afterwards it can be copied to the PYNQ board for execution and validation."
]
},
{
@@ -823,74 +821,33 @@
"metadata": {},
"outputs": [],
"source": [
- "import os\n",
+ "from shutil import copy\n",
+ "from distutils.dir_util import copy_tree\n",
"\n",
- "# set up the following values according to your own environment\n",
- "# FINN will use ssh to deploy and run the generated accelerator\n",
- "ip = \"192.168.2.99\"\n",
- "username = os.getenv(\"PYNQ_USERNAME\", \"xilinx\")\n",
- "password = os.getenv(\"PYNQ_PASSWORD\", \"xilinx\")\n",
- "port = os.getenv(\"PYNQ_PORT\", 22)\n",
- "target_dir = os.getenv(\"PYNQ_TARGET_DIR\", \"/home/xilinx/finn_tfc_end2end_example\")\n",
- "# set up ssh options to only allow publickey authentication\n",
- "options = \"-o PreferredAuthentications=publickey -o PasswordAuthentication=no\"\n",
+ "# create directory for deployment files\n",
+ "deployment_dir = make_build_dir(prefix=\"pynq_deployment_\")\n",
+ "model.set_metadata_prop(\"pynq_deployment_dir\", deployment_dir)\n",
"\n",
- "# test access to PYNQ board\n",
- "! ssh {options} {username}@{ip} -p {port} cat /var/run/motd.dynamic"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": null,
- "metadata": {},
- "outputs": [],
- "source": [
- "from finn.transformation.fpgadataflow.make_deployment import DeployToPYNQ\n",
+ "# get and copy necessary files\n",
+ "# .bit and .hwh file\n",
+ "bitfile = model.get_metadata_prop(\"bitfile\")\n",
+ "hwh_file = model.get_metadata_prop(\"hw_handoff\")\n",
+ "deploy_files = [bitfile, hwh_file]\n",
"\n",
- "model = model.transform(DeployToPYNQ(ip, port, username, password, target_dir))\n",
- "model.save(build_dir + \"/tfc_w1_a1_pynq_deploy.onnx\")"
- ]
- },
- {
- "cell_type": "markdown",
- "metadata": {},
- "source": [
- "Let's verify that the remote access credentials is saved in the model metadata, and that the deployment folder has been successfully copied to the board:"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": null,
- "metadata": {},
- "outputs": [],
- "source": [
- "model.model.metadata_props"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": null,
- "metadata": {},
- "outputs": [],
- "source": [
- "target_dir_pynq = target_dir + \"/\" + model.get_metadata_prop(\"pynq_deployment_dir\").split(\"/\")[-1]\n",
- "target_dir_pynq"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": null,
- "metadata": {},
- "outputs": [],
- "source": [
- "! ssh {options} {username}@{ip} -p {port} 'ls -l {target_dir_pynq}'"
+ "for dfile in deploy_files:\n",
+ " if dfile is not None:\n",
+ " copy(dfile, deployment_dir)\n",
+ "\n",
+ "# driver.py and python libraries\n",
+ "pynq_driver_dir = model.get_metadata_prop(\"pynq_driver_dir\")\n",
+ "copy_tree(pynq_driver_dir, deployment_dir)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
- "We only have two more steps to be able to remotely execute the deployed bitfile with some test data from the MNIST dataset. Let's load up some test data that comes bundled with FINN."
+ "Next to these files, we will also need an example numpy array to test the network on the PYNQ board. You may recall that one \"reshape\" node was left out of the StreamingDataflowPartition. We'll do that manually with a numpy function call when passing in the input, but everything else in the network ended up inside the StreamingDataflowPartition so that's all we need to do. The example numpy array can then be saved as .npy file. "
]
},
{
@@ -918,14 +875,17 @@
"iname = model.graph.input[0].name\n",
"oname = parent_model.graph.output[0].name\n",
"ishape = model.get_tensor_shape(iname)\n",
- "print(\"Expected network input shape is \" + str(ishape))"
+ "print(\"Expected network input shape is \" + str(ishape))\n",
+ "np.save(deployment_dir + \"/input.npy\", x.reshape(ishape))"
]
},
{
- "cell_type": "markdown",
+ "cell_type": "code",
+ "execution_count": null,
"metadata": {},
+ "outputs": [],
"source": [
- "Finally, we can call `execute_onnx` on the graph, which will internally call remote execution with the bitfile, grab the results and return a numpy array. You may recall that one \"reshape\" node was left out of the StreamingDataflowPartition. We'll do that manually with a numpy function call when passing in the input, but everything else in the network ended up inside the StreamingDataflowPartition so that's all we need to do."
+ "! ls {deployment_dir}"
]
},
{
@@ -934,27 +894,34 @@
"metadata": {},
"outputs": [],
"source": [
- "import numpy as np\n",
- "from finn.core.onnx_exec import execute_onnx\n",
- "\n",
- "input_dict = {iname: x.reshape(ishape)}\n",
- "ret = execute_onnx(model, input_dict)"
+ "from shutil import make_archive\n",
+ "make_archive('deploy-on-pynq-tfc', 'zip', deployment_dir)"
]
},
{
- "cell_type": "code",
- "execution_count": null,
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "You can now download the created zipfile (**File -> Open**, mark the checkbox next to the `deploy-on-pynq-tfc.zip` and select Download from the toolbar), then copy it to your PYNQ board (for instance via `scp` or `rsync`). Then, run the following commands **on the PYNQ board** to extract the archive and run the execution:"
+ ]
+ },
+ {
+ "cell_type": "markdown",
"metadata": {},
- "outputs": [],
"source": [
- "ret[oname]"
+ "```shell\n",
+ "unzip deploy-on-pynq-tfc.zip -d finn-tfc-demo\n",
+ "cd finn-tfc-demo\n",
+ "sudo python3.6 -m pip install bitstring\n",
+ "sudo python3.6 driver.py --exec_mode=execute --batchsize=1 --bitfile=resizer.bit --inputfile=input.npy\n",
+ "```"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
- "We see that the network correctly predicts this as a digit 2."
+ "The output will be saved on the PYNQ board as `output.npy` and can be copied to the host and opened with `np.load()`."
]
},
{
@@ -963,45 +930,25 @@
"source": [
"### Validating the Accuracy on a PYNQ Board <a id='validation'></a>\n",
"\n",
- "All the command line prompts here are meant to be executed with `sudo` on the PYNQ board, so we'll use a workaround (`echo password | sudo -S command`) to get that working from this notebook running on the host computer.\n",
- "\n",
"**Ensure that your PYNQ board has a working internet connecting for the next steps, since there is some downloading involved.**\n",
"\n",
"To validate the accuracy, we first need to install the [`dataset-loading`](https://github.com/fbcotter/dataset_loading) Python package to the PYNQ board. This will give us a convenient way of downloading and accessing the MNIST dataset.\n",
"\n",
"\n",
- "Command to execute on PYNQ:\n",
+ "Command to execute on PYNQ board:\n",
"\n",
"```sudo pip3 install git+https://github.com/fbcotter/dataset_loading.git@0.0.4#egg=dataset_loading```"
]
},
- {
- "cell_type": "code",
- "execution_count": null,
- "metadata": {},
- "outputs": [],
- "source": [
- "! ssh {options} -t {username}@{ip} -p {port} 'echo {password} | sudo -S pip3 install git+https://github.com/fbcotter/dataset_loading.git@0.0.4#egg=dataset_loading'"
- ]
- },
{
"cell_type": "markdown",
"metadata": {},
"source": [
"We can now use the `validate.py` script that was generated together with the driver to measure top-1 accuracy on the MNIST dataset.\n",
"\n",
- "Command to execute on PYNQ:\n",
+ "Command to execute on PYNQ board:\n",
"\n",
- "`python3.6 validate.py --dataset mnist --batchsize 1000`"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": null,
- "metadata": {},
- "outputs": [],
- "source": [
- "! ssh {options} -t {username}@{ip} -p {port} 'cd {target_dir_pynq}; echo {password} | sudo -S python3.6 validate.py --dataset mnist --batchsize 1000'"
+ "`sudo python3.6 validate.py --dataset mnist --batchsize 1000`"
]
},
{
@@ -1016,54 +963,24 @@
"metadata": {},
"source": [
"### Throughput Test on PYNQ Board <a id='throughput'></a>\n",
- "In addition to the functional verification, FINN also offers the possibility to measure the network performance directly on the PYNQ board. This can be done using the core function `throughput_test`. In the next section we import the function and execute it.\n",
- "First we extract the `remote_exec_model` again and pass it to the function. The function returns the metrics of the network as dictionary. "
- ]
- },
- {
- "cell_type": "code",
- "execution_count": null,
- "metadata": {},
- "outputs": [],
- "source": [
- "from finn.core.throughput_test import throughput_test_remote\n",
- "\n",
- "model = ModelWrapper(build_dir + \"/tfc_w1_a1_pynq_deploy.onnx\")\n",
- "res = throughput_test_remote(model, 10000)\n",
- "print(\"Network metrics:\")\n",
- "for key in res:\n",
- " print(str(key) + \": \" + str(res[key]))"
+ "In addition to the functional verification, FINN also offers the possibility to measure the network performance directly on the PYNQ board. This can be done setting the `exec_mode` to `throughput_test`. \n",
+ "Command to execute on PYNQ board:"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
- "Together with the values for folding we can evaluate the performance of our accelerator. Each layer has a total folding factor of 64 and because the network is fully pipelined, it follows: `II = 64`. II is the initiation interval and indicates how many cycles are needed for one input to be processed. "
- ]
- },
- {
- "cell_type": "code",
- "execution_count": null,
- "metadata": {},
- "outputs": [],
- "source": [
- "II = 64\n",
- "# frequency in MHz\n",
- "f_MHz = 100\n",
- "# expected throughput in MFPS\n",
- "expected_throughput = f_MHz / II\n",
- "# measured throughput (FPS) from throughput test, converted to MFPS\n",
- "measured_throughput = res[\"throughput[images/s]\"] * 0.000001\n",
- "# peformance\n",
- "print(\"We reach approximately \" + str(round((measured_throughput / expected_throughput)*100)) + \"% of the ideal performance.\")"
+ "```shell\n",
+ "sudo python3.6 driver.py --exec_mode=throughput_test --batchsize=1000 --bitfile=resizer.bit\n",
+ "```"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
- "The measured values were recorded with a batch size of 10000 and at a frequency of 100 MHz. We will be improving the efficiency of the generated accelerator examples in the coming FINN releases."
+ "The network metrics from the throughput test are saved in a file called `nw_metrics.txt` on the PYNQ board. Which can be investigated after running the command above."
]
}
],
--
GitLab