Merge pull request #104 from Xilinx/dev

Pre-release

.github/workflows/pre-commit.yml

@@ -0,0 +1,24 @@
+name: Pre-commit
+
+on:
+  pull_request:
+    branches: [ main, dev ]
+  push:
+    branches: [ main, dev ]
+
+jobs:
+  lint:
+    name: Lint PR or Push to DEV
+    runs-on: ubuntu-latest
+
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v3
+
+      - name: Setup Python
+        uses: actions/setup-python@v4
+        with:
+          python-version: '3.10'
+
+      - name: Run Lint
+        uses: pre-commit/action@v3.0.0

.gitignore

@@ -13,6 +13,8 @@ build/finn
 *.link_summary
 *.onnx
 *.pyc
+*.npz
+*.npy
 output_*/
 release/
 *.egg-info

.pre-commit-config.yaml

@@ -1,38 +1,71 @@
+# Copyright (C) 2023, Advanced Micro Devices, Inc.
+# All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions are met:
+#
+# * Redistributions of source code must retain the above copyright notice, this
+#   list of conditions and the following disclaimer.
+#
+# * Redistributions in binary form must reproduce the above copyright notice,
+#   this list of conditions and the following disclaimer in the documentation
+#   and/or other materials provided with the distribution.
+#
+# * Neither the name of FINN nor the names of its
+#   contributors may be used to endorse or promote products derived from
+#   this software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+# DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
+# FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+# DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+# SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+# CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+# OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+exclude: '^docs/conf.py'
+
 default_language_version:
-    python: python3
+    python: python3.10
 
 repos:
-- repo: git://github.com/pre-commit/pre-commit-hooks
-  rev: v3.2.0
+- repo: https://github.com/pre-commit/pre-commit-hooks
+  rev: v4.4.0
   hooks:
   - id: trailing-whitespace
+    exclude: '\.dat$'
   - id: check-added-large-files
   - id: check-ast
   - id: check-json
   - id: check-merge-conflict
   - id: check-xml
   - id: check-yaml
   - id: debug-statements
+    exclude: '^src/finn/builder/build_dataflow.py$'
   - id: end-of-file-fixer
   - id: requirements-txt-fixer
   - id: mixed-line-ending
     args: ['--fix=no']
 
-- repo: git://github.com/PyCQA/isort
-  rev: 5.5.3
+- repo: https://github.com/PyCQA/isort
+  rev: 5.12.0
   hooks:
   - id: isort
 
-- repo: git://github.com/psf/black
-  rev: stable
+- repo: https://github.com/psf/black
+  rev: 23.3.0
   hooks:
   - id: black
     language_version: python3
+    args: [--line-length=100]
 
-- repo: https://gitlab.com/pycqa/flake8
-  rev: 3.8.3
+- repo: https://github.com/PyCQA/flake8
+  rev: 6.0.0
   hooks:
   - id: flake8
     # black-compatible flake-8 config
-    args: ['--max-line-length=88',  # black default
+    args: ['--max-line-length=100',  # black default
            '--extend-ignore=E203']  # E203 is not PEP8 compliant

AMD-license-agreement-for-non-commercial-models.md

@@ -0,0 +1,14 @@
+ADVANCED MICRO DEVICES, INC.
+LICENSE AGREEMENT FOR NON-COMMERCIAL MODELS
+
+Trained Models:
+
+bincop-cnv, cnv-w1a1, cnv-w1a2, cnv-w2a2, kwsmlp-w3a3, mobilenetv1-w4a4, resnet50-w1a2, tfc-w1a1, tfc-w1a2, tfc-w2a2, unsw_nb15-mlp-w2a2, vgg10-radioml-w4a4, cnv_1w1a_gtsrb.onnx
+
+This License Agreement for Non-Commercial Models (“Agreement”) is a legal agreement between you (either an individual or an entity) and Advanced Micro Devices, Inc. on behalf of itself and its subsidiaries and affiliates (collectively “AMD”). DO NOT USE THE TRAINED MODELS IDENTIFIED ABOVE UNTIL YOU HAVE CAREFULLY READ THIS AGREEMENT. BY USING, INSTALLING, MODIFYING, COPYING, TRAINING, BENCHMARKING, OR DISTRIBUTING THE TRAINED MODELS, YOU AGREE TO AND ACCEPT ALL TERMS AND CONDITIONS OF THIS AGREEMENT. If you do not accept these terms, do not use the Trained Models.
+
+1. Subject to your compliance with this Agreement, AMD grants you a license to use, modify, and distribute the Trained Models solely for non-commercial and research purposes. This means you may use the Trained Models for benchmarking, testing, and evaluating the Trained Models (including non-commercial research undertaken by or funded by a commercial entity) but you cannot use the Trained Models in any commercial offering, including as part of a product or service you use or provide to others in exchange for money or other consideration.
+
+2. Your license to the Trained Models is subject to the following conditions: (a) you cannot alter any copyright, trademark, or other notice in the Trained Models; (b) you cannot sublicense or distribute the Trained Models under any other terms or conditions; (c) you cannot use AMD’s trademarks in your applications or technologies in a way that suggests your applications or technologies are endorsed by AMD; (d) if you distribute a Trained Model, you must provide corresponding source code for such Trained Model; and (e) if the Trained Models include any code or content subject to an open source license or third party license (“Third Party Materials”), you agree to comply with such license terms.
+
+3. THE TRAINED MODELS (INCLUDING THIRD PARTY MATERIALS, IF ANY) ARE PROVIDED “AS IS” AND WITHOUT A WARRANTY OF ANY KIND, WHETHER EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. YOU BEAR ALL RISK OF USING THE TRAINED MODELS (INCLUDING THIRD PARTY MATERIALS, IF ANY) AND YOU AGREE TO RELEASE AMD FROM ANY LIABILITY OR DAMAGES FOR ANY CLAIM OR ACTION ARISING OUT OF OR IN CONNECTION WITH YOUR USE OF THE TRAINED MODELS AND/OR THIRD PARTY MATERIALS

AUTHORS.rst

@@ -2,9 +2,18 @@
 Contributors
 ============
 
-* Yaman Umuroglu (@maltanar) (maintainer)
+* Mirza Mrahorovic (@mmrahorovic) (maintainer)
 * Jakoba Petri-Koenig (@auphelia)
+* Yaman Umuroglu (@maltanar)
 * Lucian Petrica (@quetric)
 * Tobias Alonso (@Tobi-Alonso)
 * Hendrik Borras (@HenniOVP)
 * Felix Paul Jentzsch (@felixpj)
+* Aziz Bahri (@azizb-xlnx)
+* John Monks (@jmonks-amd)
+* Fionn O'Donohoe (@fionnodonohoe-xlnx)
+* Radoslav Pitoňák (@rpitonak)
+* Matthias Gehre (@mgehre-amd)
+* NaelF (@NaelF)
+* Tim Paine (@timkpaine)
+* Hannah Yan (@hannahxy13)

LICENSE

@@ -1,20 +1,22 @@
 BSD 3-Clause License
 
-Copyright (c) 2023, Xilinx
+Copyright (C) 2020-2022, Xilinx, Inc.
+Copyright (C) 2022-2024, Advanced Micro Devices, Inc.
+All rights reserved.
 
 Redistribution and use in source and binary forms, with or without
 modification, are permitted provided that the following conditions are met:
 
-1. Redistributions of source code must retain the above copyright notice, this
-   list of conditions and the following disclaimer.
+* Redistributions of source code must retain the above copyright notice, this
+  list of conditions and the following disclaimer.
 
-2. Redistributions in binary form must reproduce the above copyright notice,
-   this list of conditions and the following disclaimer in the documentation
-   and/or other materials provided with the distribution.
+* Redistributions in binary form must reproduce the above copyright notice,
+  this list of conditions and the following disclaimer in the documentation
+  and/or other materials provided with the distribution.
 
-3. Neither the name of the copyright holder nor the names of its
-   contributors may be used to endorse or promote products derived from
-   this software without specific prior written permission.
+* Neither the name of FINN nor the names of its
+  contributors may be used to endorse or promote products derived from
+  this software without specific prior written permission.
 
 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
 AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE

README.md

@@ -5,13 +5,13 @@
         <img src="https://img.shields.io/github/v/release/Xilinx/finn-examples?color=%09%23228B22&display_name=tag&label=Release" />
     </a>
     <a href="https://github.com/Xilinx/finn/tree/v0.9">
-        <img src="https://img.shields.io/badge/FINN-v0.9.0-blue" />
+        <img src="https://img.shields.io/badge/FINN-v0.10-blue" />
     </a>
     <a href="https://github.com/Xilinx/PYNQ/tree/v3.0.1">
         <img src="https://img.shields.io/badge/PYNQ-v3.0.1-blue" />
     </a>
     <a href="https://www.xilinx.com/support/download/index.html/content/xilinx/en/downloadNav/vivado-design-tools/2022-1.html">
-        <img src="https://img.shields.io/badge/Vivado%2FVitis-v2022.1-blue" />
+        <img src="https://img.shields.io/badge/Vivado%2FVitis-v2022.2-blue" />
     </a>
 </p>
 
@@ -51,7 +51,7 @@ on your PYNQ board:
 python3 -m pip install pip==23.0 setuptools==67.1.0
 ```
 
-Since we are going to install finn-examples without build-isolation, we need to ensure all dependencies are installed. For that, install `setuptools_csm` as well:
+Since we are going to install finn-examples without build-isolation, we need to ensure all dependencies are installed. For that, install `setuptools_scm` as well:
 
 ```shell
 python3 -m pip install setuptools_scm==7.1.0
@@ -124,18 +124,18 @@ dummy_out = accel.execute(dummy_in)
 
 ##  Example Neural Network Accelerators
 | Dataset                                                        | Topology                | Quantization                                               | Supported boards | Supported build flows
-|----------------------------------------------------------------|-------------------------|------------------------------------------------------------|------------------|------------------|
-| <img src="docs/img/cifar-10.png" width="150"/><br/>CIFAR-10     | CNV (VGG-11-like)       | several variants:<br>1/2-bit weights/activations           | Pynq-Z1<br>ZCU104<br>Ultra96<br>U250              | Pynq-Z1<br>ZCU104<br>Ultra96<br>U250 |
-| <img src="docs/img/mnist.jpg" width="150"/><br/><br>MNIST       | 3-layer fully-connected | several variants:<br>1/2-bit weights/activations           | Pynq-Z1<br>ZCU104<br>Ultra96<br>U250              | Pynq-Z1<br>ZCU104<br>Ultra96<br>U250 |
-| <img src="docs/img/imagenet.jpg" width="150"/><br/><br>ImageNet | MobileNet-v1            | 4-bit weights & activations<br>8-bit first layer weights | Alveo U250       | Alveo U250 |
-| <img src="docs/img/imagenet.jpg" width="150"/><br/><br>ImageNet | ResNet-50            | 1-bit weights 2-bit activations<br>4-bit residuals<br>8-bit first/last layer weights | Alveo U250       | - |
-| <img src="docs/img/radioml.png" width="150"/><br/><br>RadioML 2018 | 1D CNN (VGG10)     |  4-bit weights & activations | ZCU104  | ZCU104 |
-| <img src="docs/img/maskedfacenet.jpg" width="150"/><br/><br>MaskedFace-Net | [BinaryCoP](https://arxiv.org/pdf/2102.03456)<br/>*Contributed by TU Munich+BMW*  | 1-bit weights & activations | Pynq-Z1       | Pynq-Z1 |
-| <img src="docs/img/keyword-spotting.png" width="150"/><br/><br>Google Speech Commands v2 | 3-layer fully-connected  | 3-bit weights & activations | Pynq-Z1       | Pynq-Z1 |
-| <img src="docs/img/unsw-nb15.jpg" width="150"/><br/><br>UNSW-NB15 | 4-layer fully-connected  | 2-bit weights & activations | Pynq-Z1 <br> ZCU104 <br> Ultra96       | Pynq-Z1 <br> ZCU104 <br> Ultra96 |
-
-*Please note that the build flow for ResNet-50 for the Alveo U250 has known issues and we're currently working on resolving them. However, you can still execute the associated notebook, as we provide a pre-built FPGA bitfile generated with an older Vivado (/FINN) version targeting the [xilinx_u250_xdma_201830_2](https://www.xilinx.com/products/boards-and-kits/alveo/package-files-archive/u250-2018-3-1.html) platform.* <br>
-*Furthermore, please note that you can target other boards (such as the Pynq-Z2 or ZCU102) by changing the build script manually, but these accelerators have not been tested.*
+|:----------------------------------------------------------------:|:-------------------------:|:------------------------------------------------------------:|:------------------:|:------------------:|
+| CIFAR-10     | CNV (VGG-11-like)       | several variants:<br>1/2-bit weights/activations           | Pynq-Z1<br>ZCU104<br>Ultra96<br>U250              | Pynq-Z1<br>ZCU104<br>Ultra96<br>U250 |
+| MNIST       | 3-layer fully-connected | several variants:<br>1/2-bit weights/activations           | Pynq-Z1<br>ZCU104<br>Ultra96<br>U250              | Pynq-Z1<br>ZCU104<br>Ultra96<br>U250 |
+| ImageNet | MobileNet-v1            | 4-bit weights & activations<br>8-bit first layer weights | Alveo U250       | Alveo U250 |
+| ImageNet | ResNet-50            | 1-bit weights 2-bit activations<br>4-bit residuals<br>8-bit first/last layer weights | Alveo U250       | Alveo U250 |
+| RadioML 2018 | 1D CNN (VGG10)     |  4-bit weights & activations | ZCU104  | ZCU104 |
+| MaskedFace-Net | [BinaryCoP](https://arxiv.org/pdf/2102.03456)<br/>*Contributed by TU Munich+BMW*  | 1-bit weights & activations | Pynq-Z1       | Pynq-Z1 |
+| Google Speech Commands v2 | 3-layer fully-connected  | 3-bit weights & activations | Pynq-Z1       | Pynq-Z1 |
+| UNSW-NB15 | 4-layer fully-connected  | 2-bit weights & activations | Pynq-Z1 <br> ZCU104 <br> Ultra96       | Pynq-Z1 <br> ZCU104 <br> Ultra96 |
+| GTSRB | CNV (VGG-11-like) | 1-bit weights & activations | Pynq-Z1 | Pynq-Z1 |
+
+*Please note that you can target other boards (such as the Pynq-Z2 or ZCU102) by changing the build script manually, but these accelerators have not been tested.*
 
 We welcome community contributions to add more examples to this repo!
 

build/README.md

@@ -16,7 +16,7 @@ to do this again in the future when the `finn-examples` repo gets updated and re
 Docker community edition `docker-ce`.
 
 3. Set up the environment variables to point to your Vivado/Vitis installation, depending on your target platform(s):
-    *  For Zynq platforms you'll need to set `VIVADO_PATH`, e.g. `VIVADO_PATH=/opt/xilinx/Vivado/2019.1/`
+    * For Zynq platforms you'll need to set `VIVADO_PATH`, e.g. `VIVADO_PATH=/opt/xilinx/Vivado/2022.2/`
     * For Alveo platforms you'll need to set `VITIS_PATH`, `PLATFORM_REPO_PATHS` and `XILINX_XRT`
 
 ## Build bitfiles

build/bnn-pynq/README.md

@@ -42,7 +42,7 @@ cd $FINN_EXAMPLES/build/finn
 The BNN-PYNQ networks are part of the
 [Brevitas examples](https://github.com/Xilinx/brevitas/tree/master/src/brevitas_examples/bnn_pynq). You can find the details on quantization, accuracy, layers used in the Brevitas repo, as well as the training scripts if you'd like to retrain them yourself.
 
-Subsequently, those trained networks are [exported to ONNX](https://github.com/Xilinx/finn/blob/master/notebooks/basics/1_brevitas_network_import.ipynb). In addition, the particular versions
+Subsequently, those trained networks are [exported to ONNX](https://github.com/Xilinx/finn/blob/main/notebooks/basics/1_brevitas_network_import_via_QONNX.ipynb). In addition, the particular versions
 used here have two additions, as described in the "Adding Pre- and Postprocessing" section of [this notebook](https://github.com/Xilinx/finn/blob/master/notebooks/end2end_example/bnn-pynq/tfc_end2end_example.ipynb):
 
 * A divide-by-255 node is added at the input, and the input is marked as 8-bit (to directly accept 8-bit images as input)

build/bnn-pynq/build.py

@@ -1,3 +1,31 @@
+# Copyright (C) 2024, Advanced Micro Devices, Inc.
+# All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions are met:
+#
+# * Redistributions of source code must retain the above copyright notice, this
+#   list of conditions and the following disclaimer.
+#
+# * Redistributions in binary form must reproduce the above copyright notice,
+#   this list of conditions and the following disclaimer in the documentation
+#   and/or other materials provided with the distribution.
+#
+# * Neither the name of FINN nor the names of its
+#   contributors may be used to endorse or promote products derived from
+#   this software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+# DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
+# FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+# DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+# SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+# CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+# OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
 import finn.builder.build_dataflow as build
 import finn.builder.build_dataflow_config as build_cfg
 from finn.util.basic import alveo_default_platform
@@ -54,12 +82,15 @@ def platform_to_shell(platform):
         cfg = build_cfg.DataflowBuildConfig(
             output_dir="output_%s_%s" % (model_name, release_platform_name),
             folding_config_file="folding_config/%s_folding_config.json" % model_name,
-            synth_clk_period_ns=10.0,
+            synth_clk_period_ns=5.0,
             board=platform_name,
             shell_flow_type=shell_flow_type,
             vitis_platform=vitis_platform,
             generate_outputs=[build_cfg.DataflowOutputType.BITFILE],
             save_intermediate_models=True,
+            default_swg_exception=True,
+            specialize_layers_config_file="specialize_layers_config/%s_specialize_layers.json"
+            % model_name,
         )
         model_file = "models/%s.onnx" % model_name
         # launch FINN compiler to build