ONNX Hub Docs (#3712)

* ONNX Hub Docs
Signed-off-by: NMark <mhamilton723@gmail.com>

* Finish sections
Signed-off-by: NMark <mhamilton723@gmail.com>

docs/Hub.md

+<!--- SPDX-License-Identifier: Apache-2.0 -->
+
+# ONNX Model Hub
+
+The ONNX Model Hub is a simple and fast way to get started with state of the art pre-trained 
+ONNX models from the [ONNX Model Zoo](https://github.com/onnx/models). Furthermore, this allows researchers and model 
+developers the opportunity to share their pre-trained models with the broader community.
+
+## Install
+ The ONNX Model hub will be included in the `onnx` package from version 1.11 onwards. 
+ To use the hub before the 1.11 release please install from the weekly build:
+ 
+```shell script
+pip install -i https://test.pypi.org/simple/ onnx-weekly 
+```
+
+## Basic usage
+The ONNX Model Hub is capable of downloading, listing, and querying trained models from any git repository,
+ and defaults to the official [ONNX Model Zoo](https://github.com/onnx/models). In this section we demonstrate some of the basic functionality. 
+
+First please import the hub using:
+```python
+from onnx import hub
+```
+
+#### Downloading a model by name:
+
+The `load` function will default to searching the model zoo for the latest model with a matching name,
+ download this model to a local cache, and load the model into a `ModelProto` 
+ object for use with the ONNX runtime.
+
+```python
+model = hub.load("resnet50")
+```
+
+
+#### Downloading from custom repositories:
+
+Any repository with the proper structure can be a ONNX model hub. To download from other hubs,
+ or to specify a particular branch or commit on the main model hub one can provide the `repo` parameter:
+
+```python
+model = hub.load("resnet50", repo='onnx/models:771185265efbdc049fb223bd68ab1aeb1aecde76')
+```
+
+#### Listing and inspecting Models:
+
+The model hub provides APIs for querying the model zoo to learn more about available models.
+ This does not download the models, but rather just returns information about models matching the given arguments
+ 
+ ```python
+# List all models in the onnx/models:master repo
+all_models = hub.list_models()
+
+# List all versions/opsets of a specific model
+mnist_models = hub.list_models(model="mnist")
+
+# List all models matching a given "tag"
+vision_models = hub.list_models(tags=["vision"])
+```
+
+One can also inspect the metadata of a model prior to download with the `get_model_info` function:
+
+```python
+print(hub.get_model_info(model="mnist", opset=8))
+```
+This will print something like:
+```
+ModelInfo(
+    model=MNIST,
+    opset=8,
+    path=vision/classification/mnist/model/mnist-8.onnx,
+    metadata={
+     'model_sha': '2f06e72de813a8635c9bc0397ac447a601bdbfa7df4bebc278723b958831c9bf',
+     'model_bytes': 26454,
+     'tags': ['vision', 'classification', 'mnist'],
+     'io_ports': {
+        'inputs': [{'name': 'Input3', 'shape': [1, 1, 28, 28], 'type': 'tensor(float)'}],
+        'outputs': [{'name': 'Plus214_Output_0', 'shape': [1, 10], 'type': 'tensor(float)'}]},
+     'model_with_data_path': 'vision/classification/mnist/model/mnist-8.tar.gz',
+     'model_with_data_sha': '1dd098b0fe8bc750585eefc02013c37be1a1cae2bdba0191ccdb8e8518b3a882',
+     'model_with_data_bytes': 25962}
+)
+```
+
+
+## Local Caching
+
+The ONNX Model hub locally caches downloaded models in a configurable location 
+so that subsequent calls to `hub.load` do not require network connection.
+
+#### Default cache location
+The hub client looks for the following default cache locations in this order:
+
+1) `$ONNX_HOME/hub` if the `ONNX_HOME` environment variable is defined
+2) `$XDG_CACHE_HOME/hub` if the `XDG_CACHE_HOME` environment variable is defined
+3) `~/.cache/onnx/hub` where `~` is the user home directory
+
+#### Setting the cache location
+To manually set the cache location use:
+
+```python
+hub.set_dir("my/cache/directory")
+```
+
+Additionally one can inspect the cache location with:
+
+```python
+print(hub.get_dir())
+```
+
+#### Additional cache details
+
+To clear the model cache one can simply delete the cache directory using a python utility like `shutil` or `os`. 
+Furthermore one can choose to override the cached model using the `force_reload` option:
+
+```python
+model = hub.load("resnet50", force_reload=True)
+```
+
+We include this flag for completeness but note that models in the cache are disambiguated with sha256 hashes so 
+ the force_reload flag is not necessary for normal use.
+Finally we note that the model cache directory structure will mirror the directory structure 
+specified by the `model_path` field of the manifest, but with file names disambiguated with model SHA256 Hashes. 
+
+This way, the model cache is human readable, can disambiguate between multiple versions of models,
+ and can re-use cached models across different hubs if they have the same name and hash. 
+
+## Architecture
+
+![ONNX Hub Architecture](images/onnx_hub_arch.svg)
+
+The ONNX Hub consists of two main components, the client and the server.
+ The client code currently is included in the `onnx` package and can be pointed at a
+  server in the form of a hosted `ONNX_HUB_MANIFEST.json` within a github repository 
+  such as [the one in the ONNX Model Zoo](https://github.com/onnx/models/blob/master/ONNX_HUB_MANIFEST.json).
+  This manifest file is a JSON document which lists all models and their metadata
+   and is designed to be programming language agnostic. An example of a well formed model manifest entry is as follows:
+   
+   ```json
+{
+    "model": "BERT-Squad",
+    "model_path": "text/machine_comprehension/bert-squad/model/bertsquad-8.onnx",
+    "onnx_version": "1.3",
+    "opset_version": 8,
+    "metadata": {
+        "model_sha": "cad65b9807a5e0393e4f84331f9a0c5c844d9cc736e39781a80f9c48ca39447c",
+        "model_bytes": 435882893,
+        "tags": ["text", "machine comprehension", "bert-squad"],
+        "io_ports": {
+            "inputs": [
+                {
+                    "name": "unique_ids_raw_output___9:0",
+                    "shape": ["unk__475"],
+                    "type": "tensor(int64)"
+                },
+                {
+                    "name": "segment_ids:0",
+                    "shape": ["unk__476", 256],
+                    "type": "tensor(int64)"
+                },
+                {
+                    "name": "input_mask:0",
+                    "shape": ["unk__477", 256],
+                    "type": "tensor(int64)"
+                },
+                {
+                    "name": "input_ids:0",
+                    "shape": ["unk__478", 256],
+                    "type": "tensor(int64)"
+                }
+            ],
+            "outputs": [
+                {
+                    "name": "unstack:1",
+                    "shape": ["unk__479", 256],
+                    "type": "tensor(float)"
+                },
+                {
+                    "name": "unstack:0",
+                    "shape": ["unk__480", 256],
+                    "type": "tensor(float)"
+                },
+                {
+                    "name": "unique_ids:0",
+                    "shape": ["unk__481"],
+                    "type": "tensor(int64)"
+                }
+            ]
+        },
+        "model_with_data_path": "text/machine_comprehension/bert-squad/model/bertsquad-8.tar.gz",
+        "model_with_data_sha": "c8c6c7e0ab9e1333b86e8415a9d990b2570f9374f80be1c1cb72f182d266f666",
+        "model_with_data_bytes": 403400046
+    }
+}
+```
+These important fields are:
+
+- `model`: The name of the model used for querying
+- `model_path`: The relative path of the model stored in Git LFS. 
+- `onnx_version`: The ONNX version of the model
+- `opset_version`: The version of the opset. The client downloads the latest opset if left unspecified.
+- `metadata/model_sha`: Optional model sha specification for increased download security
+- `metadata/tags`: Optional high level tags to help users find models by a given type
+
+All other fields in the `metadata` field are optional for the client but provide important details for users.
+
+## Adding to the ONNX Model Hub
+
+
+#### Contributing an official model
+
+The simplest way to add a model to the official `onnx/models` version model hub is to follow 
+[these guidelines](https://github.com/onnx/models/blob/master/contribute.md) to contribute your model. Once contributed,
+ensure that your model has a markdown table in its `README.md` 
+([Example](https://github.com/onnx/models/tree/master/vision/classification/mobilenet)). The model hub
+ manifest generator will pull information from these markdown tables. To run the generator:
+ 
+ ```shell script
+git clone https://github.com/onnx/models.git
+git lfs pull --include="*" --exclude=""
+cd models/workflow_scripts
+python generate_onnx_hub_manifest.py
+```
+
+Once a new manifest is generated add, submit it in a pull request to ``onnx/models``
+
+#### Hosting your own ONNX Model Hub
+
+To host your own model hub, add an `ONNX_HUB_MANIFEST.json` to the top level of your github repository
+ ([Example](https://github.com/onnx/models/blob/master/ONNX_HUB_MANIFEST.json)). At a minimum your 
+ manifest entries should include the fields mentioned in 
+ the [Architecture Section](Hub.md#Architecture) of this document.
+ Once committed, check that you can download models 
+ using the "Downloading from custom repositories" section of this doc.
\ No newline at end of file

docs/images/onnx_hub_arch.svg

+<svg width="3126" height="2133" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" overflow="hidden"><defs><clipPath id="clip0"><rect x="541" y="263" width="3126" height="2133"/></clipPath></defs><g clip-path="url(#clip0)" transform="translate(-541 -263)"><rect x="541" y="272" width="3126" height="2124"/><path d="M2442.5 738.837C2442.5 646.971 2516.97 572.5 2608.84 572.5L3471.16 572.5C3563.03 572.5 3637.5 646.971 3637.5 738.837L3637.5 1404.16C3637.5 1496.03 3563.03 1570.5 3471.16 1570.5L2608.84 1570.5C2516.97 1570.5 2442.5 1496.03 2442.5 1404.16Z" stroke="#FFFFFF" stroke-width="4.58333" stroke-miterlimit="8" fill="#404040" fill-rule="evenodd"/><text fill="#FFFFFF" font-family="Calibri,Calibri_MSFontService,sans-serif" font-weight="400" font-size="83" transform="matrix(1 0 0 1 2789.98 714)">Git LFS Storage<tspan font-family="Calibri Light,Calibri Light_MSFontService,sans-serif" font-weight="300" font-size="202" x="-1743.23" y="-201">ONNX Hub Architecture</tspan></text><path d="M2534.5 817.667C2534.5 802.663 2546.66 790.5 2561.67 790.5L2842.33 790.5C2857.34 790.5 2869.5 802.663 2869.5 817.667L2869.5 926.333C2869.5 941.337 2857.34 953.5 2842.33 953.5L2561.67 953.5C2546.66 953.5 2534.5 941.337 2534.5 926.333Z" stroke="#2F528F" stroke-width="4.58333" stroke-miterlimit="8" fill="#4472C4" fill-rule="evenodd"/><text fill="#FFFFFF" font-family="Calibri,Calibri_MSFontService,sans-serif" font-weight="400" font-size="83" transform="matrix(1 0 0 1 2596.49 899)">M1 v2</text><path d="M2534.5 1023.67C2534.5 1008.66 2546.66 996.5 2561.67 996.5L2842.33 996.5C2857.34 996.5 2869.5 1008.66 2869.5 1023.67L2869.5 1132.33C2869.5 1147.34 2857.34 1159.5 2842.33 1159.5L2561.67 1159.5C2546.66 1159.5 2534.5 1147.34 2534.5 1132.33Z" stroke="#2F528F" stroke-width="4.58333" stroke-miterlimit="8" fill="#4472C4" fill-rule="evenodd"/><text fill="#FFFFFF" font-family="Calibri,Calibri_MSFontService,sans-serif" font-weight="400" font-size="83" transform="matrix(1 0 0 1 2596.49 1105)">M2 v3</text><path d="M2534.5 1379.67C2534.5 1364.66 2546.66 1352.5 2561.67 1352.5L2842.33 1352.5C2857.34 1352.5 2869.5 1364.66 2869.5 1379.67L2869.5 1488.33C2869.5 1503.34 2857.34 1515.5 2842.33 1515.5L2561.67 1515.5C2546.66 1515.5 2534.5 1503.34 2534.5 1488.33Z" stroke="#2F528F" stroke-width="4.58333" stroke-miterlimit="8" fill="#4472C4" fill-rule="evenodd"/><text fill="#FFFFFF" font-family="Calibri,Calibri_MSFontService,sans-serif" font-weight="400" font-size="83" transform="matrix(1 0 0 1 2590.76 1461)">MN v1</text><path d="M2899.5 817.667C2899.5 802.663 2911.66 790.5 2926.67 790.5L3207.33 790.5C3222.34 790.5 3234.5 802.663 3234.5 817.667L3234.5 926.333C3234.5 941.337 3222.34 953.5 3207.33 953.5L2926.67 953.5C2911.66 953.5 2899.5 941.337 2899.5 926.333Z" stroke="#2F528F" stroke-width="4.58333" stroke-miterlimit="8" fill="#4472C4" fill-rule="evenodd"/><text fill="#FFFFFF" font-family="Calibri,Calibri_MSFontService,sans-serif" font-weight="400" font-size="83" transform="matrix(1 0 0 1 2961.64 899)">M1 v1</text><path d="M2899.5 1022.83C2899.5 1007.74 2911.74 995.5 2926.83 995.5L3207.17 995.5C3222.26 995.5 3234.5 1007.74 3234.5 1022.83L3234.5 1132.17C3234.5 1147.26 3222.26 1159.5 3207.17 1159.5L2926.83 1159.5C2911.74 1159.5 2899.5 1147.26 2899.5 1132.17Z" stroke="#2F528F" stroke-width="4.58333" stroke-miterlimit="8" fill="#4472C4" fill-rule="evenodd"/><text fill="#FFFFFF" font-family="Calibri,Calibri_MSFontService,sans-serif" font-weight="400" font-size="83" transform="matrix(1 0 0 1 2961.64 1104)">M1 v2</text><path d="M3264.5 1022.83C3264.5 1007.74 3276.74 995.5 3291.83 995.5L3572.17 995.5C3587.26 995.5 3599.5 1007.74 3599.5 1022.83L3599.5 1132.17C3599.5 1147.26 3587.26 1159.5 3572.17 1159.5L3291.83 1159.5C3276.74 1159.5 3264.5 1147.26 3264.5 1132.17Z" stroke="#2F528F" stroke-width="4.58333" stroke-miterlimit="8" fill="#4472C4" fill-rule="evenodd"/><text fill="#FFFFFF" font-family="Calibri,Calibri_MSFontService,sans-serif" font-weight="400" font-size="83" transform="matrix(1 0 0 1 3326.79 1104)">M1 v2<tspan font-size="133" x="-682.733" y="175">…</tspan></text><path d="M1400.5 736.17C1400.5 645.778 1473.78 572.5 1564.17 572.5L2218.83 572.5C2309.22 572.5 2382.5 645.778 2382.5 736.17L2382.5 1406.83C2382.5 1497.22 2309.22 1570.5 2218.83 1570.5L1564.17 1570.5C1473.78 1570.5 1400.5 1497.22 1400.5 1406.83Z" stroke="#FFFFFF" stroke-width="4.58333" stroke-miterlimit="8" fill="#ED7D31" fill-rule="evenodd"/><text fill="#FFFFFF" font-family="Calibri,Calibri_MSFontService,sans-serif" font-weight="400" font-size="83" transform="matrix(1 0 0 1 1574.57 714)">onnx<tspan font-size="83" x="165.39" y="0">/models Repo</tspan></text><path d="M1524.67 1515.5C1509.66 1515.5 1497.5 1503.34 1497.5 1488.33L1497.5 817.667C1497.5 802.663 1509.66 790.5 1524.67 790.5L1633.33 790.5C1648.34 790.5 1660.5 802.663 1660.5 817.667L1660.5 1488.33C1660.5 1503.34 1648.34 1515.5 1633.33 1515.5Z" stroke="#2F528F" stroke-width="4.58333" stroke-miterlimit="8" fill="#4472C4" fill-rule="evenodd"/><text fill="#FFFFFF" font-family="Calibri,Calibri_MSFontService,sans-serif" font-weight="400" font-size="83" transform="matrix(-1.83697e-16 -1 1 -1.83697e-16 1606.11 1401)">MANIFEST.json</text><path d="M1757.5 817.667C1757.5 802.663 1769.66 790.5 1784.67 790.5L2293.33 790.5C2308.34 790.5 2320.5 802.663 2320.5 817.667L2320.5 926.333C2320.5 941.337 2308.34 953.5 2293.33 953.5L1784.67 953.5C1769.66 953.5 1757.5 941.337 1757.5 926.333Z" stroke="#2F528F" stroke-width="4.58333" stroke-miterlimit="8" fill="#4472C4" fill-rule="evenodd"/><text fill="#FFFFFF" font-family="Calibri,Calibri_MSFontService,sans-serif" font-weight="400" font-size="83" transform="matrix(1 0 0 1 1802.57 899)">M1 v2 Pointer</text><path d="M1757.5 1023.67C1757.5 1008.66 1769.66 996.5 1784.67 996.5L2293.33 996.5C2308.34 996.5 2320.5 1008.66 2320.5 1023.67L2320.5 1132.33C2320.5 1147.34 2308.34 1159.5 2293.33 1159.5L1784.67 1159.5C1769.66 1159.5 1757.5 1147.34 1757.5 1132.33Z" stroke="#2F528F" stroke-width="4.58333" stroke-miterlimit="8" fill="#4472C4" fill-rule="evenodd"/><text fill="#FFFFFF" font-family="Calibri,Calibri_MSFontService,sans-serif" font-weight="400" font-size="83" transform="matrix(1 0 0 1 1802.57 1105)">M2 v3 Pointer</text><path d="M1757.5 1379.67C1757.5 1364.66 1769.66 1352.5 1784.67 1352.5L2293.33 1352.5C2308.34 1352.5 2320.5 1364.66 2320.5 1379.67L2320.5 1488.33C2320.5 1503.34 2308.34 1515.5 2293.33 1515.5L1784.67 1515.5C1769.66 1515.5 1757.5 1503.34 1757.5 1488.33Z" stroke="#2F528F" stroke-width="4.58333" stroke-miterlimit="8" fill="#4472C4" fill-rule="evenodd"/><text fill="#FFFFFF" font-family="Calibri,Calibri_MSFontService,sans-serif" font-weight="400" font-size="83" transform="matrix(1 0 0 1 1927.54 1412)">MN v1 <tspan font-size="83" x="-10.3812" y="99">Pointer</tspan></text><path d="M2320 865.125 2499.84 865.125 2499.84 878.875 2320 878.875ZM2492.96 851.375 2534.21 872 2492.96 892.625Z" fill="#FFFFFF"/><path d="M2320 1070.12 2499.84 1070.13 2499.84 1083.88 2320 1083.88ZM2492.96 1056.38 2534.21 1077 2492.96 1097.63Z" fill="#FFFFFF"/><path d="M2320 1427.13 2499.84 1427.13 2499.84 1440.88 2320 1440.88ZM2492.96 1413.38 2534.21 1434 2492.96 1454.63Z" fill="#FFFFFF"/><path d="M1666.5 1150.77 1751.73 1399.25 1738.72 1403.71 1653.5 1155.23ZM1762.51 1388.29 1756.38 1434 1723.49 1401.67Z" fill="#FFFFFF"/><path d="M4.22417-5.42421 73.482 48.5112 65.0336 59.3596-4.22417 5.42421ZM76.5061 33.4386 96.3788 75.0562 51.1611 65.9838Z" fill="#FFFFFF" transform="matrix(1 0 0 -1 1660 1152.06)"/><path d="M6.50311-2.2305 91.7294 246.25 78.7232 250.711-6.50311 2.2305ZM102.505 235.286 96.3788 280.996 63.4865 248.669Z" fill="#FFFFFF" transform="matrix(1 0 0 -1 1660 1153)"/><path d="M603.5 695.336C603.5 627.495 658.496 572.5 726.336 572.5L1217.66 572.5C1285.5 572.5 1340.5 627.495 1340.5 695.336L1340.5 1447.66C1340.5 1515.5 1285.5 1570.5 1217.66 1570.5L726.336 1570.5C658.496 1570.5 603.5 1515.5 603.5 1447.66Z" stroke="#FFFFFF" stroke-width="4.58333" stroke-miterlimit="8" fill="#ED7D31" fill-rule="evenodd"/><text fill="#FFFFFF" font-family="Calibri,Calibri_MSFontService,sans-serif" font-weight="400" font-size="83" transform="matrix(1 0 0 1 695.493 702)">onnx<tspan font-size="83" x="165.39" y="0">/</tspan><tspan font-size="83" x="196.052" y="0">onnx</tspan><tspan font-size="83" x="380.348" y="0">Repo</tspan></text><path d="M663.5 854.834C663.5 833.11 681.11 815.5 702.834 815.5L1237.17 815.5C1258.89 815.5 1276.5 833.11 1276.5 854.834L1276.5 1012.17C1276.5 1033.89 1258.89 1051.5 1237.17 1051.5L702.834 1051.5C681.11 1051.5 663.5 1033.89 663.5 1012.17Z" stroke="#2F528F" stroke-width="4.58333" stroke-miterlimit="8" fill="#4472C4" fill-rule="evenodd"/><text fill="#FFFFFF" font-family="Calibri,Calibri_MSFontService,sans-serif" font-weight="400" font-size="83" transform="matrix(1 0 0 1 789.11 911)">ONNX Hub <tspan font-size="83" x="-44.7562" y="99">Python Client</tspan></text><path d="M603.5 1831C603.5 1774.39 649.391 1728.5 706 1728.5L3535 1728.5C3591.61 1728.5 3637.5 1774.39 3637.5 1831L3637.5 2241C3637.5 2297.61 3591.61 2343.5 3535 2343.5L706 2343.5C649.391 2343.5 603.5 2297.61 603.5 2241Z" stroke="#FFFFFF" stroke-width="4.58333" stroke-miterlimit="8" fill="#7030A0" fill-rule="evenodd"/><text fill="#FFFFFF" font-family="Calibri,Calibri_MSFontService,sans-serif" font-weight="400" font-size="83" transform="matrix(1 0 0 1 1879.05 1852)">Local Machine</text><path d="M667.5 1966.17C667.5 1933.21 694.214 1906.5 727.168 1906.5L1959.83 1906.5C1992.79 1906.5 2019.5 1933.21 2019.5 1966.17L2019.5 2204.83C2019.5 2237.79 1992.79 2264.5 1959.83 2264.5L727.168 2264.5C694.214 2264.5 667.5 2237.79 667.5 2204.83Z" stroke="#507E32" stroke-width="4.58333" stroke-miterlimit="8" fill="#70AD47" fill-rule="evenodd"/><text fill="#FFFFFF" font-family="Calibri,Calibri_MSFontService,sans-serif" font-weight="400" font-size="83" transform="matrix(1 0 0 1 1121.52 2017)">Local Process<tspan font-family="Consolas,Consolas_MSFontService,sans-serif" font-size="69" x="-398.747" y="81">import </tspan><tspan font-family="Consolas,Consolas_MSFontService,sans-serif" font-size="69" x="-134.059" y="81">onnx</tspan><tspan font-family="Consolas,Consolas_MSFontService,sans-serif" font-size="69" x="-398.747" y="169">model = </tspan><tspan font-family="Consolas,Consolas_MSFontService,sans-serif" font-size="69" x="-96.247" y="169">onnx.hub.load</tspan><tspan font-family="Consolas,Consolas_MSFontService,sans-serif" font-size="69" x="395.316" y="169">(‘</tspan><tspan font-family="Consolas,Consolas_MSFontService,sans-serif" font-size="69" x="470.941" y="169">modelN</tspan><tspan font-family="Consolas,Consolas_MSFontService,sans-serif" font-size="69" x="697.816" y="169">’)</tspan></text><path d="M2248.5 1963.17C2248.5 1930.21 2275.21 1903.5 2308.17 1903.5L3539.83 1903.5C3572.79 1903.5 3599.5 1930.21 3599.5 1963.17L3599.5 2201.83C3599.5 2234.79 3572.79 2261.5 3539.83 2261.5L2308.17 2261.5C2275.21 2261.5 2248.5 2234.79 2248.5 2201.83Z" stroke="#507E32" stroke-width="4.58333" stroke-miterlimit="8" fill="#70AD47" fill-rule="evenodd"/><text fill="#FFFFFF" font-family="Calibri,Calibri_MSFontService,sans-serif" font-weight="400" font-size="83" transform="matrix(1 0 0 1 2797.99 2014)">Model Cache</text><path d="M2535.5 2063.83C2535.5 2048.74 2547.74 2036.5 2562.83 2036.5L2843.17 2036.5C2858.26 2036.5 2870.5 2048.74 2870.5 2063.83L2870.5 2173.17C2870.5 2188.26 2858.26 2200.5 2843.17 2200.5L2562.83 2200.5C2547.74 2200.5 2535.5 2188.26 2535.5 2173.17Z" stroke="#2F528F" stroke-width="4.58333" stroke-miterlimit="8" fill="#4472C4" fill-rule="evenodd"/><text fill="#FFFFFF" font-family="Calibri,Calibri_MSFontService,sans-serif" font-weight="400" font-size="83" transform="matrix(1 0 0 1 2591.27 2145)">MN v1</text><path d="M0.0237067-6.87496 480.989-5.21646 480.942 8.53346-0.0237067 6.87496ZM474.162-18.9901 515.341 1.77703 474.02 22.2597Z" fill="#FFFFFF" transform="matrix(-1 0 0 1 2534.34 2118)"/><path d="M2708.87 1514.99 2709.36 2001.07 2695.61 2001.08 2695.12 1515.01ZM2723.1 1994.18 2702.52 2035.45 2681.85 1994.22Z" fill="#FFFFFF"/><path d="M946.875 1050 946.876 1871.31 933.126 1871.31 933.125 1050ZM960.626 1864.44 940.001 1905.69 919.376 1864.44Z" fill="#FFFFFF"/><path d="M1585.87 1515 1585.88 1870.83 1572.13 1870.83 1572.12 1515ZM1599.63 1863.96 1579 1905.21 1558.38 1863.96Z" fill="#FFFFFF"/></g></svg>
\ No newline at end of file