update service data

.gitignore

@@ -0,0 +1,3 @@
+.DS_Store
+__pycache__
+model

README.md

@@ -1,3 +1,35 @@
-# test_mlflow
+# Abstract
 
-測試使用MLflow紀錄Pytorch模型訓練，以及從Model registry中拉取Production model進行推論。
+Try to use [MLflow](https://mlflow.org) platform to log PyTorch model training, and pull production model from model registry to run inference⛩
+
+# Requirements
+
+* MacOS 12.5
+* Docker 20.10
+
+# Dirs
+
+* **service**
+  * House MLflow service data, including MLflow artifacts, backend store and model registry
+* **env**
+  * **mlflow.yaml**
+    * conda env yaml to run this repo
+
+# Files
+
+* **docker-compose.yaml**
+  * a yaml to apply docker-compose to start MLflow service with basic configuration (run ```docker-compose -f docker-compose.yaml up```)
+* **test_pytorch_m1.py**
+  * a script to test PyTorch on Apple M1 platform with GPU acceleration
+* **train.py**
+  * a sample code to apply PyTorch to train a small neural network to predict fortune with MLflow logging
+* **predict.py**
+  * a sample code to call registered model to predict testing data and save model to local file system
+* **get_registered_model_via_rest_api.py**
+  * a script to test MLflow REST api
+* **log_unsupported_model.py**
+  * a sample script to apply mlflow.pyfunc to package unsupported ml model which can be logged and registered by mlflow
+* **optimize_model.py**
+  * a sample script to demonstrate how to use MLflow and TensorRT libs to optimize Pytorch model on edge devices and fetch it out on client
+
+###### tags: `MLOps`

artifacts/0/24469fc083d6470a9cad7f17a6eeeea0/artifacts/model/MLmodel

@@ -1,16 +0,0 @@
-artifact_path: model
-flavors:
-  python_function:
-    data: data
-    env: conda.yaml
-    loader_module: mlflow.pytorch
-    pickle_module_name: mlflow.pytorch.pickle_module
-    python_version: 3.10.9
-  pytorch:
-    code: null
-    model_data: data
-    pytorch_version: 1.13.1
-mlflow_version: 1.30.0
-model_uuid: 2382b7a39c064e7b9b1465cfd84140a3
-run_id: 24469fc083d6470a9cad7f17a6eeeea0
-utc_time_created: '2023-02-21 05:57:41.973454'

artifacts/0/24469fc083d6470a9cad7f17a6eeeea0/artifacts/model/conda.yaml

@@ -1,11 +0,0 @@
-channels:
-- conda-forge
-dependencies:
-- python=3.10.9
-- pip<=23.0.1
-- pip:
-  - mlflow
-  - cloudpickle==2.2.1
-  - torch==1.13.1
-  - tqdm==4.64.1
-name: mlflow-env

artifacts/0/24469fc083d6470a9cad7f17a6eeeea0/artifacts/model/data/pickle_module_info.txt

@@ -1 +0,0 @@
-mlflow.pytorch.pickle_module

artifacts/0/24469fc083d6470a9cad7f17a6eeeea0/artifacts/model/python_env.yaml

@@ -1,7 +0,0 @@
-python: 3.10.9
-build_dependencies:
-- pip==23.0.1
-- setuptools==67.3.2
-- wheel==0.38.4
-dependencies:
-- -r requirements.txt

artifacts/0/24469fc083d6470a9cad7f17a6eeeea0/artifacts/model/requirements.txt

@@ -1,4 +0,0 @@
-mlflow
-cloudpickle==2.2.1
-torch==1.13.1
-tqdm==4.64.1

artifacts/0/2820b379bfc945358bfd516e5577846c/artifacts/cls_model/MLmodel

@@ -1,16 +0,0 @@
-artifact_path: cls_model
-flavors:
-  python_function:
-    data: data
-    env: conda.yaml
-    loader_module: mlflow.pytorch
-    pickle_module_name: mlflow.pytorch.pickle_module
-    python_version: 3.10.9
-  pytorch:
-    code: null
-    model_data: data
-    pytorch_version: 1.13.1
-mlflow_version: 1.30.0
-model_uuid: e40643f3e1b9481896e1ae6ed30e8654
-run_id: 2820b379bfc945358bfd516e5577846c
-utc_time_created: '2023-02-21 05:33:10.779919'

artifacts/0/2820b379bfc945358bfd516e5577846c/artifacts/cls_model/conda.yaml

@@ -1,10 +0,0 @@
-channels:
-- conda-forge
-dependencies:
-- python=3.10.9
-- pip<=23.0.1
-- pip:
-  - mlflow
-  - cloudpickle==2.2.1
-  - torch==1.13.1
-name: mlflow-env

artifacts/0/2820b379bfc945358bfd516e5577846c/artifacts/cls_model/data/pickle_module_info.txt

@@ -1 +0,0 @@
-mlflow.pytorch.pickle_module

artifacts/0/2820b379bfc945358bfd516e5577846c/artifacts/cls_model/python_env.yaml

@@ -1,7 +0,0 @@
-python: 3.10.9
-build_dependencies:
-- pip==23.0.1
-- setuptools==67.3.2
-- wheel==0.38.4
-dependencies:
-- -r requirements.txt

artifacts/0/2820b379bfc945358bfd516e5577846c/artifacts/cls_model/requirements.txt

@@ -1,3 +0,0 @@
-mlflow
-cloudpickle==2.2.1
-torch==1.13.1

artifacts/0/3ef01a1e3e3d4ba2be705da789bbb8e1/artifacts/models/MLmodel

@@ -1,16 +0,0 @@
-artifact_path: models
-flavors:
-  python_function:
-    data: data
-    env: conda.yaml
-    loader_module: mlflow.pytorch
-    pickle_module_name: mlflow.pytorch.pickle_module
-    python_version: 3.10.9
-  pytorch:
-    code: null
-    model_data: data
-    pytorch_version: 1.13.1
-mlflow_version: 1.30.0
-model_uuid: faf1bec9ecb64581b22a0b8e09a9cca8
-run_id: 3ef01a1e3e3d4ba2be705da789bbb8e1
-utc_time_created: '2023-02-21 05:07:17.344052'

artifacts/0/3ef01a1e3e3d4ba2be705da789bbb8e1/artifacts/models/conda.yaml

@@ -1,10 +0,0 @@
-channels:
-- conda-forge
-dependencies:
-- python=3.10.9
-- pip<=23.0.1
-- pip:
-  - mlflow
-  - cloudpickle==2.2.1
-  - torch==1.13.1
-name: mlflow-env

artifacts/0/3ef01a1e3e3d4ba2be705da789bbb8e1/artifacts/models/data/pickle_module_info.txt

@@ -1 +0,0 @@
-mlflow.pytorch.pickle_module

artifacts/0/3ef01a1e3e3d4ba2be705da789bbb8e1/artifacts/models/python_env.yaml

@@ -1,7 +0,0 @@
-python: 3.10.9
-build_dependencies:
-- pip==23.0.1
-- setuptools==67.3.2
-- wheel==0.38.4
-dependencies:
-- -r requirements.txt

artifacts/0/3ef01a1e3e3d4ba2be705da789bbb8e1/artifacts/models/requirements.txt

@@ -1,3 +0,0 @@
-mlflow
-cloudpickle==2.2.1
-torch==1.13.1

artifacts/0/63c7363339e042f4848d9041ba8deb82/artifacts/cls_model/MLmodel

@@ -1,16 +0,0 @@
-artifact_path: cls_model
-flavors:
-  python_function:
-    data: data
-    env: conda.yaml
-    loader_module: mlflow.pytorch
-    pickle_module_name: mlflow.pytorch.pickle_module
-    python_version: 3.10.9
-  pytorch:
-    code: null
-    model_data: data
-    pytorch_version: 1.13.1
-mlflow_version: 1.30.0
-model_uuid: a0ecc970cadb47a9b839283e9514732d
-run_id: 63c7363339e042f4848d9041ba8deb82
-utc_time_created: '2023-02-21 05:37:55.904472'

artifacts/0/63c7363339e042f4848d9041ba8deb82/artifacts/cls_model/conda.yaml

@@ -1,10 +0,0 @@
-channels:
-- conda-forge
-dependencies:
-- python=3.10.9
-- pip<=23.0.1
-- pip:
-  - mlflow
-  - cloudpickle==2.2.1
-  - torch==1.13.1
-name: mlflow-env

artifacts/0/63c7363339e042f4848d9041ba8deb82/artifacts/cls_model/data/pickle_module_info.txt

@@ -1 +0,0 @@
-mlflow.pytorch.pickle_module

artifacts/0/63c7363339e042f4848d9041ba8deb82/artifacts/cls_model/python_env.yaml

@@ -1,7 +0,0 @@
-python: 3.10.9
-build_dependencies:
-- pip==23.0.1
-- setuptools==67.3.2
-- wheel==0.38.4
-dependencies:
-- -r requirements.txt

artifacts/0/63c7363339e042f4848d9041ba8deb82/artifacts/cls_model/requirements.txt

@@ -1,3 +0,0 @@
-mlflow
-cloudpickle==2.2.1
-torch==1.13.1

artifacts/0/6845ef0d54024cb3bdb32050f6a46fea/artifacts/models/MLmodel

@@ -1,16 +0,0 @@
-artifact_path: models
-flavors:
-  python_function:
-    data: data
-    env: conda.yaml
-    loader_module: mlflow.pytorch
-    pickle_module_name: mlflow.pytorch.pickle_module
-    python_version: 3.10.9
-  pytorch:
-    code: null
-    model_data: data
-    pytorch_version: 1.13.1
-mlflow_version: 1.30.0
-model_uuid: 4ebe94bd0249452a90b3497d3b00a1c3
-run_id: 6845ef0d54024cb3bdb32050f6a46fea
-utc_time_created: '2023-02-21 05:25:14.020335'

artifacts/0/6845ef0d54024cb3bdb32050f6a46fea/artifacts/models/conda.yaml

@@ -1,10 +0,0 @@
-channels:
-- conda-forge
-dependencies:
-- python=3.10.9
-- pip<=23.0.1
-- pip:
-  - mlflow
-  - cloudpickle==2.2.1
-  - torch==1.13.1
-name: mlflow-env

artifacts/0/6845ef0d54024cb3bdb32050f6a46fea/artifacts/models/data/pickle_module_info.txt

@@ -1 +0,0 @@
-mlflow.pytorch.pickle_module

artifacts/0/6845ef0d54024cb3bdb32050f6a46fea/artifacts/models/python_env.yaml

@@ -1,7 +0,0 @@
-python: 3.10.9
-build_dependencies:
-- pip==23.0.1
-- setuptools==67.3.2
-- wheel==0.38.4
-dependencies:
-- -r requirements.txt

artifacts/0/6845ef0d54024cb3bdb32050f6a46fea/artifacts/models/requirements.txt

@@ -1,3 +0,0 @@
-mlflow
-cloudpickle==2.2.1
-torch==1.13.1

artifacts/0/68e8a3cbbafa46538ebb8a60d80f185d/artifacts/models/MLmodel

@@ -1,16 +0,0 @@
-artifact_path: models
-flavors:
-  python_function:
-    data: data
-    env: conda.yaml
-    loader_module: mlflow.pytorch
-    pickle_module_name: mlflow.pytorch.pickle_module
-    python_version: 3.10.9
-  pytorch:
-    code: null
-    model_data: data
-    pytorch_version: 1.13.1
-mlflow_version: 1.30.0
-model_uuid: 8cd1e70114e548ea8d9bfb1bf468e285
-run_id: 68e8a3cbbafa46538ebb8a60d80f185d
-utc_time_created: '2023-02-21 05:05:46.624814'

artifacts/0/68e8a3cbbafa46538ebb8a60d80f185d/artifacts/models/conda.yaml

@@ -1,10 +0,0 @@
-channels:
-- conda-forge
-dependencies:
-- python=3.10.9
-- pip<=23.0.1
-- pip:
-  - mlflow
-  - cloudpickle==2.2.1
-  - torch==1.13.1
-name: mlflow-env

artifacts/0/68e8a3cbbafa46538ebb8a60d80f185d/artifacts/models/data/pickle_module_info.txt

@@ -1 +0,0 @@
-mlflow.pytorch.pickle_module

artifacts/0/68e8a3cbbafa46538ebb8a60d80f185d/artifacts/models/python_env.yaml

@@ -1,7 +0,0 @@
-python: 3.10.9
-build_dependencies:
-- pip==23.0.1
-- setuptools==67.3.2
-- wheel==0.38.4
-dependencies:
-- -r requirements.txt

artifacts/0/68e8a3cbbafa46538ebb8a60d80f185d/artifacts/models/requirements.txt

@@ -1,3 +0,0 @@
-mlflow
-cloudpickle==2.2.1
-torch==1.13.1

artifacts/0/8ba27f225a7442be8816977c2077c510/artifacts/models/MLmodel

@@ -1,16 +0,0 @@
-artifact_path: models
-flavors:
-  python_function:
-    data: data
-    env: conda.yaml
-    loader_module: mlflow.pytorch
-    pickle_module_name: mlflow.pytorch.pickle_module
-    python_version: 3.10.9
-  pytorch:
-    code: null
-    model_data: data
-    pytorch_version: 1.13.1
-mlflow_version: 1.30.0
-model_uuid: dd1b1e3a6b5f4274a5690a8843751ff3
-run_id: 8ba27f225a7442be8816977c2077c510
-utc_time_created: '2023-02-21 05:05:04.660670'

artifacts/0/8ba27f225a7442be8816977c2077c510/artifacts/models/conda.yaml

@@ -1,10 +0,0 @@
-channels:
-- conda-forge
-dependencies:
-- python=3.10.9
-- pip<=23.0.1
-- pip:
-  - mlflow
-  - cloudpickle==2.2.1
-  - torch==1.13.1
-name: mlflow-env

artifacts/0/8ba27f225a7442be8816977c2077c510/artifacts/models/data/pickle_module_info.txt

@@ -1 +0,0 @@
-mlflow.pytorch.pickle_module

artifacts/0/8ba27f225a7442be8816977c2077c510/artifacts/models/python_env.yaml

@@ -1,7 +0,0 @@
-python: 3.10.9
-build_dependencies:
-- pip==23.0.1
-- setuptools==67.3.2
-- wheel==0.38.4
-dependencies:
-- -r requirements.txt

artifacts/0/8ba27f225a7442be8816977c2077c510/artifacts/models/requirements.txt

@@ -1,3 +0,0 @@
-mlflow
-cloudpickle==2.2.1
-torch==1.13.1

artifacts/0/f1320882f24c4f489cbf85159627eaf8/artifacts/cls_model/MLmodel

@@ -1,16 +0,0 @@
-artifact_path: cls_model
-flavors:
-  python_function:
-    data: data
-    env: conda.yaml
-    loader_module: mlflow.pytorch
-    pickle_module_name: mlflow.pytorch.pickle_module
-    python_version: 3.10.9
-  pytorch:
-    code: null
-    model_data: data
-    pytorch_version: 1.13.1
-mlflow_version: 1.30.0
-model_uuid: aaa800b217da4dd0b8f17e8dbfdc5c45
-run_id: f1320882f24c4f489cbf85159627eaf8
-utc_time_created: '2023-02-21 05:34:08.242864'

artifacts/0/f1320882f24c4f489cbf85159627eaf8/artifacts/cls_model/conda.yaml

@@ -1,10 +0,0 @@
-channels:
-- conda-forge
-dependencies:
-- python=3.10.9
-- pip<=23.0.1
-- pip:
-  - mlflow
-  - cloudpickle==2.2.1
-  - torch==1.13.1
-name: mlflow-env

artifacts/0/f1320882f24c4f489cbf85159627eaf8/artifacts/cls_model/data/pickle_module_info.txt

@@ -1 +0,0 @@
-mlflow.pytorch.pickle_module

artifacts/0/f1320882f24c4f489cbf85159627eaf8/artifacts/cls_model/python_env.yaml

@@ -1,7 +0,0 @@
-python: 3.10.9
-build_dependencies:
-- pip==23.0.1
-- setuptools==67.3.2
-- wheel==0.38.4
-dependencies:
-- -r requirements.txt

artifacts/0/f1320882f24c4f489cbf85159627eaf8/artifacts/cls_model/requirements.txt

@@ -1,3 +0,0 @@
-mlflow
-cloudpickle==2.2.1
-torch==1.13.1

docker-compose.yaml

@@ -0,0 +1,18 @@
+version: '3.7'
+
+services:
+  mlflow_server:
+    image: ghcr.io/mlflow/mlflow:v2.1.1
+    restart: always
+    ports:
+      - 5001:5001
+    volumes:
+      - ~/python/test_mlflow/service:/home
+    healthcheck:
+      test: ["CMD", "curl", "-f", "http://0.0.0.0:5001"]
+      interval: 30s
+      timeout: 10s
+      retries: 3
+      start_period: 20s
+    command: bash -c "apt update && apt install -y curl && mlflow server --host 0.0.0.0 --port 5001 --backend-store-uri sqlite:////home/backend.db --registry-store-uri sqlite:////home/registry.db --artifacts-destination /home/artifacts --serve-artifacts"
+

env/mlflow.yaml

@@ -1,15 +1,31 @@
-name: torch
+name: mlflow
 channels:
   - pytorch
   - conda-forge
 dependencies:
   - alembic=1.9.4
   - aom=3.5.0
-  - appdirs=1.4.4
+  - arrow-cpp=9.0.0
+  - aws-c-auth=0.6.24
+  - aws-c-cal=0.5.20
+  - aws-c-common=0.8.11
+  - aws-c-compression=0.2.16
+  - aws-c-event-stream=0.2.18
+  - aws-c-http=0.7.4
+  - aws-c-io=0.13.17
+  - aws-c-mqtt=0.8.6
+  - aws-c-s3=0.2.4
+  - aws-c-sdkutils=0.1.7
+  - aws-checksums=0.1.14
+  - aws-crt-cpp=0.19.7
+  - aws-sdk-cpp=1.10.57
   - bcrypt=3.2.2
   - blinker=1.5
+  - brotli=1.0.9
+  - brotli-bin=1.0.9
   - brotlipy=0.7.0
   - bzip2=1.0.8
+  - c-ares=1.18.1
   - ca-certificates=2022.12.7
   - certifi=2022.12.7
   - cffi=1.15.1
@@ -18,7 +34,9 @@ dependencies:
   - cloudpickle=2.2.1
   - colorama=0.4.6
   - configparser=5.3.0
+  - contourpy=1.0.7
   - cryptography=39.0.1
+  - cycler=0.11.0
   - databricks-cli=0.17.4
   - docker-py=6.0.0
   - entrypoints=0.4
@@ -32,10 +50,13 @@ dependencies:
   - fontconfig=2.14.2
   - fonts-conda-ecosystem=1
   - fonts-conda-forge=1
+  - fonttools=4.38.0
   - freetype=2.12.1
   - gettext=0.21.1
+  - gflags=2.2.2
   - gitdb=4.0.10
   - gitpython=3.1.31
+  - glog=0.6.0
   - gmp=6.2.1
   - gnutls=3.7.8
   - greenlet=2.0.2
@@ -46,56 +67,85 @@ dependencies:
   - importlib_resources=5.12.0
   - itsdangerous=2.1.2
   - jinja2=3.1.2
+  - joblib=1.2.0
   - jpeg=9e
+  - kiwisolver=1.4.4
+  - krb5=1.20.1
   - lame=3.100
   - lcms2=2.14
   - lerc=4.0.0
+  - libabseil=20230125.0
   - libblas=3.9.0
+  - libbrotlicommon=1.0.9
+  - libbrotlidec=1.0.9
+  - libbrotlienc=1.0.9
   - libcblas=3.9.0
-  - libcxx=14.0.6
+  - libcrc32c=1.1.2
+  - libcurl=7.88.1
+  - libcxx=15.0.7
   - libdeflate=1.17
+  - libedit=3.1.20191231
+  - libev=4.33
+  - libevent=2.1.10
   - libffi=3.4.2
   - libgfortran=5.0.0
   - libgfortran5=11.3.0
+  - libgoogle-cloud=2.7.0
+  - libgrpc=1.51.1
   - libiconv=1.17
   - libidn2=2.3.4
   - liblapack=3.9.0
+  - libllvm11=11.1.0
+  - libnghttp2=1.51.0
   - libopenblas=0.3.21
   - libopus=1.3.1
   - libpng=1.6.39
   - libprotobuf=3.21.12
   - libsodium=1.0.18
   - libsqlite=3.40.0
+  - libssh2=1.10.0
   - libtasn1=4.19.0
+  - libthrift=0.18.0
   - libtiff=4.5.0
   - libunistring=0.9.10
+  - libutf8proc=2.8.0
   - libvpx=1.11.0
   - libwebp-base=1.2.4
   - libxcb=1.13
   - libxml2=2.10.3
   - libzlib=1.2.13
   - llvm-openmp=15.0.7
+  - llvmlite=0.39.1
+  - lz4-c=1.9.4
   - mako=1.2.4
+  - markdown=3.4.1
   - markupsafe=2.1.2
-  - mlflow=1.30.0
+  - matplotlib-base=3.7.0
+  - mlflow=2.1.1
+  - munkres=1.1.4
   - ncurses=6.3
   - nettle=3.8.1
-  - numpy=1.24.2
+  - numba=0.56.4
+  - numpy=1.23.5
   - oauthlib=3.2.2
   - openh264=2.3.1
   - openjpeg=2.5.0
   - openssl=3.0.8
+  - orc=1.8.2
   - p11-kit=0.24.1
-  - packaging=21.3
+  - packaging=22.0
   - pandas=1.5.3
   - paramiko=3.0.0
+  - parquet-cpp=1.5.1
   - pillow=9.4.0
   - pip=23.0.1
-  - pooch=1.6.0
+  - platformdirs=3.0.0
+  - pooch=1.7.0
   - prometheus_client=0.16.0
-  - prometheus_flask_exporter=0.22.0
+  - prometheus_flask_exporter=0.22.2
   - protobuf=4.21.12
   - pthread-stubs=0.4
+  - pyarrow=9.0.0
   - pycparser=2.21
   - pyjwt=2.6.0
   - pynacl=1.5.0
@@ -110,22 +160,30 @@ dependencies:
   - pywin32-on-windows=0.1.0
   - pyyaml=6.0
   - querystring_parser=1.2.4
+  - re2=2023.02.02
   - readline=8.1.2
   - requests=2.28.2
-  - scipy=1.10.0
-  - setuptools=67.3.2
+  - scikit-learn=1.2.1
+  - scipy=1.10.1
+  - setuptools=67.4.0
+  - shap=0.41.0
   - six=1.16.0
+  - slicer=0.0.7
   - smmap=3.0.5
+  - snappy=1.1.9
   - sqlalchemy=1.4.46
   - sqlparse=0.4.3
   - svt-av1=1.4.1
   - tabulate=0.9.0
+  - threadpoolctl=3.1.0
   - tk=8.6.12
   - torchaudio=0.13.1
   - torchvision=0.14.1
   - tqdm=4.64.1
+  - typing-extensions=4.4.0
   - typing_extensions=4.4.0
   - tzdata=2022g
+  - unicodedata2=15.0.0
   - urllib3=1.26.14
   - websocket-client=1.5.1
   - werkzeug=2.2.3
@@ -136,6 +194,7 @@ dependencies:
   - xorg-libxdmcp=1.1.3
   - xz=5.2.6
   - yaml=0.2.5
-  - zipp=3.14.0
+  - zipp=3.15.0
+  - zlib=1.2.13
   - zstd=1.5.2
-prefix: /Users/xiao_deng/miniforge3/envs/torch
+prefix: /Users/xiao_deng/miniforge3/envs/mlflow

get_registered_model_via_rest_api.py

@@ -0,0 +1,30 @@
+# get_registered_model_via_rest_api.py
+#
+# author: deng
+# date  : 20230224
+
+import json
+import requests
+
+def main():
+
+    registered_model_name = 'fortune_predict_model'
+    production_model_version = None
+
+    query = {'name': registered_model_name}
+    res = requests.get('http://127.0.0.1:5001/api/2.0/mlflow/registered-models/get', params=query)
+    content = json.loads(res.text)
+    print(content)
+
+    for model in content['registered_model']['latest_versions']:
+
+        if model['current_stage'] == 'Production':
+            production_model_version = model['version']
+
+    if production_model_version is not None:
+        query = {'name': registered_model_name, 'version': production_model_version}
+    res = requests.get('http://127.0.0.1:5001/api/2.0/mlflow/model-versions/get-download-uri', params=query)
+    print(res.text)
+
+if __name__ == '__main__':
+    main()

log_unsupported_model.py

@@ -0,0 +1,77 @@
+# log_unsupported_model.py
+#
+# author: deng
+# date  : 20230309
+
+import mlflow
+import pandas as pd
+
+
+class CustomModel(mlflow.pyfunc.PythonModel):
+    """ A mlflow wrapper to package unsupported model """
+
+    def __init__(self):
+        self.model = None
+
+    def load_model(self):
+        # load your custom model here
+        self.model = lambda value: value * 2
+
+    def predict(self,
+                model_input: pd.DataFrame) -> pd.DataFrame:
+
+        if self.model is None:
+            self.load_model()
+
+        output = model_input.apply(self.model)
+
+        return output
+
+
+def log_model(server_uri:str,
+              exp_name: str,
+              registered_model_name: str) -> None:
+
+    # init mlflow
+    mlflow.set_tracking_uri(server_uri)
+    mlflow.set_experiment(exp_name)
+
+    # register custom model
+    model = CustomModel()
+    mlflow.pyfunc.log_model(
+        artifact_path='model',
+        python_model=model,
+        registered_model_name=registered_model_name
+    )
+
+
+def pull_model(server_uri: str,
+               exp_name: str,
+               registered_model_name: str) -> None:
+
+    # init mlflow
+    mlflow.set_tracking_uri(server_uri)
+    mlflow.set_experiment(exp_name)
+
+    # pull model from registry
+    model = mlflow.pyfunc.load_model(f'models:/{registered_model_name}/latest')
+    model = model.unwrap_python_model()  # get CustomModel object
+    print(f'Model loaded. model type: {type(model)}')
+
+    # test model availability
+    fake_data = pd.DataFrame([1, 3, 5])
+    output = model.predict(fake_data)
+    print(f'input data: {fake_data}, predictions: {output}')
+
+    # save it to local file system
+    mlflow.pyfunc.save_model(path=f'./model/{exp_name}', python_model=model)
+
+
+if __name__ == '__main__':
+
+    server_uri = 'http://127.0.0.1:5001'
+    exp_name = 'custom_model'
+    registered_model_name = 'custom_model'
+
+    log_model(server_uri, exp_name, registered_model_name)
+    pull_model(server_uri, exp_name, registered_model_name)

optimize_model.py

@@ -0,0 +1,75 @@
+# optimize_model.py
+#
+# author: deng
+# date  : 20230418
+
+import shutil
+from pathlib import Path
+
+import torch
+import mlflow
+
+
+def optimize_pytorch_model(run_id: str,
+                           model_artifact_dir: str = 'model') -> None:
+    """Optimize Pytorch model from MLflow server on edge device
+
+    Args:
+        run_id (str): mlflow run id
+        model_artifact_dir (str, optional): model dir of run on server. Defaults to 'model'.
+    """
+
+    download_path = Path('./model/downloaded_pytorch_model')
+    if download_path.is_dir():
+        print(f'Remove existed dir: {download_path}')
+        shutil.rmtree(download_path)
+
+    # Download model artifacts to local file system
+    mlflow_model = mlflow.pytorch.load_model(Path(f'runs:/{run_id}').joinpath(model_artifact_dir).as_posix())
+    mlflow.pytorch.save_model(mlflow_model, download_path)
+
+    # Optimize model
+    model = torch.load(download_path.joinpath('data/model.pth'))
+    dummy_input = torch.randn(5)
+    torch.onnx.export(model, dummy_input, download_path.joinpath('data/model.onnx'))
+    # we can not call TensorRT on macOS, so imagine we get a serialized model😘
+    download_path.joinpath('data/model.trt').touch()
+
+    # Sent optimized model back to given run
+    with mlflow.start_run(run_id=run_id):
+        mlflow.log_artifact(download_path.joinpath('data/model.trt'), 'model/data')
+        print(f'Optimized model had been uploaded to server: {mlflow.get_tracking_uri()}')
+
+
+def download_optimized_model(run_id: str,
+                             save_dir: str,
+                             model_artifact_path: str = 'model/data/model.trt') -> None:
+    """Download optimized model from MLflow server on clent
+
+    Args:
+        run_id (str): mlflow run id
+        save_dir (str): dir of local file system to save model
+        model_artifact_path (str, optional): artifact path of model on server. Defaults to 'model/data/model.trt'.
+    """
+
+    mlflow.artifacts.download_artifacts(
+        run_id= run_id,
+        artifact_path=model_artifact_path,
+        dst_path=save_dir
+    )
+
+    print(f'Optimized model had been saved, please check: {Path(save_dir).joinpath(model_artifact_path)}')
+
+
+if __name__ == '__main__':
+
+    mlflow.set_tracking_uri('http://127.0.0.1:5001')
+
+    optimize_pytorch_model(
+        run_id='f1b7b9a5ba934f158c07975a8a332de5'
+    )
+
+    download_optimized_model(
+        run_id='f1b7b9a5ba934f158c07975a8a332de5',
+        save_dir='./model/download_tensorrt'
+    )

predict.py

@@ -7,15 +7,22 @@ import torch
 import mlflow
 
 
-if __name__ == '__main__':
+def main():
 
     # set MLflow server
-    mlflow.set_tracking_uri('http://127.0.0.1:5000')
+    mlflow.set_tracking_uri('http://127.0.0.1:5001')
 
     # load production model
-    model = mlflow.pytorch.load_model('models:/cls_model/production')
+    model = mlflow.pytorch.load_model('models:/fortune_predict_model/production')
 
     # predict
-    fake_data = torch.randn(10)
-    output = model(fake_data)
-    print(output)
+    my_personal_info = torch.randn(5)
+    my_fortune = model(my_personal_info)
+    print(my_fortune)
+
+    # save model and env to local file system
+    mlflow.pytorch.save_model(model, './model/fortune_predict_model')
+
+
+if __name__ == '__main__':
+    main()

service/artifacts/3/f1b7b9a5ba934f158c07975a8a332de5/artifacts/code/train.py

@@ -0,0 +1,98 @@
+# train.py
+#
+# author: deng
+# date  : 20230221
+
+import torch
+import torch.nn as nn
+from torch.optim import SGD
+import mlflow
+from mlflow.models.signature import ModelSignature
+from mlflow.types.schema import Schema, ColSpec
+from tqdm import tqdm
+
+
+class Net(nn.Module):
+    """ define a simple neural network model """
+    def __init__(self):
+        super(Net, self).__init__()
+        self.fc1 = nn.Linear(5, 3)
+        self.fc2 = nn.Linear(3, 1)
+
+    def forward(self, x):
+        x = self.fc1(x)
+        x = torch.relu(x)
+        x = self.fc2(x)
+        return x
+
+
+def train(model, dataloader, criterion, optimizer, epochs):
+    """ define the training function """
+    for epoch in tqdm(range(epochs), 'Epochs'):
+
+        for batch, (inputs, labels) in enumerate(dataloader):
+
+            # forwarding
+            outputs = model(inputs)
+            loss = criterion(outputs, labels)
+
+            # update gradient
+            optimizer.zero_grad()
+            loss.backward()
+            optimizer.step()
+
+        # log loss
+        mlflow.log_metric('train_loss', loss.item(), step=epoch)
+
+    return loss
+
+
+if __name__ == '__main__':
+
+    # set hyper parameters
+    learning_rate = 1e-2
+    batch_size = 10
+    epochs = 20
+
+    # create a dataloader with fake data
+    dataloader = [(torch.randn(5), torch.randn(1)) for _ in range(100)]
+    dataloader = torch.utils.data.DataLoader(dataloader, batch_size=batch_size)
+
+    # create the model, criterion, and optimizer
+    model = Net()
+    criterion = nn.MSELoss()
+    optimizer = SGD(model.parameters(), lr=learning_rate)
+
+    # set the tracking URI to the model registry
+    mlflow.set_tracking_uri('http://127.0.0.1:5001')
+    mlflow.set_experiment('train_fortune_predict_model')
+
+    # start a new MLflow run
+    with mlflow.start_run():
+
+        # train the model
+        loss = train(model, dataloader, criterion, optimizer, epochs)
+
+        # log some additional metrics
+        mlflow.log_metric('final_loss', loss.item())
+        mlflow.log_param('learning_rate', learning_rate)
+        mlflow.log_param('batch_size', batch_size)
+
+        # create a signature to record model input and output info
+        input_schema = Schema([
+            ColSpec('float', 'age'),
+            ColSpec('float', 'mood level'),
+            ColSpec('float', 'health level'),
+            ColSpec('float', 'hungry level'),
+            ColSpec('float', 'sexy level')
+        ])
+        output_schema = Schema([ColSpec('float', 'fortune')])
+        signature = ModelSignature(inputs=input_schema, outputs=output_schema)
+
+        # log trained model
+        mlflow.pytorch.log_model(model, 'model', signature=signature)
+
+        # log training code
+        mlflow.log_artifact('./train.py', 'code')
+
+    print('Completed.')

start_mlflow_server.sh

@@ -1,7 +0,0 @@
-#!/bin/bash
-# start_mlflow_server.sh
-#
-# author: deng
-# date  : 20230221
-
-mlflow server --backend-store-uri sqlite:///mlflow.db --default-artifact-root ./artifacts

test_pytorch_m1.py

@@ -0,0 +1,14 @@
+# test_pytorch_m1.py
+# Ref: https://towardsdatascience.com/installing-pytorch-on-apple-m1-chip-with-gpu-acceleration-3351dc44d67c
+#
+# author: deng
+# date  : 20230301
+
+import torch
+import math
+
+print('This ensures that the current MacOS version is at least 12.3+')
+print(torch.backends.mps.is_available())
+
+print('\nThis ensures that the current current PyTorch installation was built with MPS activated.')
+print(torch.backends.mps.is_built())

train.py

@@ -6,16 +6,18 @@
 import torch
 import torch.nn as nn
 from torch.optim import SGD
-from tqdm import tqdm
 import mlflow
+from mlflow.models.signature import ModelSignature
+from mlflow.types.schema import Schema, ColSpec
+from tqdm import tqdm
 
 
 class Net(nn.Module):
     """ define a simple neural network model """
     def __init__(self):
         super(Net, self).__init__()
-        self.fc1 = nn.Linear(10, 5)
-        self.fc2 = nn.Linear(5, 1)
+        self.fc1 = nn.Linear(5, 3)
+        self.fc2 = nn.Linear(3, 1)
 
     def forward(self, x):
         x = self.fc1(x)
@@ -28,7 +30,7 @@ def train(model, dataloader, criterion, optimizer, epochs):
     """ define the training function """
     for epoch in tqdm(range(epochs), 'Epochs'):
 
-        for i, (inputs, labels) in enumerate(dataloader):
+        for batch, (inputs, labels) in enumerate(dataloader):
 
             # forwarding
             outputs = model(inputs)
@@ -40,7 +42,7 @@ def train(model, dataloader, criterion, optimizer, epochs):
             optimizer.step()
 
         # log loss
-            mlflow.log_metric('train_loss', loss.item(), step=i)
+        mlflow.log_metric('train_loss', loss.item(), step=epoch)
 
     return loss
 
@@ -49,11 +51,12 @@ if __name__ == '__main__':
 
     # set hyper parameters
     learning_rate = 1e-2
+    batch_size = 10
     epochs = 20
 
     # create a dataloader with fake data
-    dataloader = [(torch.randn(10), torch.randn(1)) for _ in range(100)]
-    dataloader = torch.utils.data.DataLoader(dataloader, batch_size=10)
+    dataloader = [(torch.randn(5), torch.randn(1)) for _ in range(100)]
+    dataloader = torch.utils.data.DataLoader(dataloader, batch_size=batch_size)
 
     # create the model, criterion, and optimizer
     model = Net()
@@ -61,19 +64,35 @@ if __name__ == '__main__':
     optimizer = SGD(model.parameters(), lr=learning_rate)
 
     # set the tracking URI to the model registry
-    mlflow.set_tracking_uri('http://127.0.0.1:5000')
+    mlflow.set_tracking_uri('http://127.0.0.1:5001')
+    mlflow.set_experiment('train_fortune_predict_model')
 
-    # start the MLflow run
+    # start a new MLflow run
     with mlflow.start_run():
 
-        # train the model and log the loss
+        # train the model
         loss = train(model, dataloader, criterion, optimizer, epochs)
 
         # log some additional metrics
         mlflow.log_metric('final_loss', loss.item())
         mlflow.log_param('learning_rate', learning_rate)
+        mlflow.log_param('batch_size', batch_size)
+
+        # create a signature to record model input and output info
+        input_schema = Schema([
+            ColSpec('float', 'age'),
+            ColSpec('float', 'mood level'),
+            ColSpec('float', 'health level'),
+            ColSpec('float', 'hungry level'),
+            ColSpec('float', 'sexy level')
+        ])
+        output_schema = Schema([ColSpec('float', 'fortune')])
+        signature = ModelSignature(inputs=input_schema, outputs=output_schema)
 
         # log trained model
-        mlflow.pytorch.log_model(model, 'model')
+        mlflow.pytorch.log_model(model, 'model', signature=signature)
+
+        # log training code
+        mlflow.log_artifact('./train.py', 'code')
 
     print('Completed.')