Logistic Regression #3329
Logistic Regression #3329
Conversation
Logisitic Regression trained on iris dataset
iris_pipeline_project/Dockerfile
Outdated
| ENV PYTHONPATH=/ovms/lib/python | ||
|
|
||
| RUN apt update && apt install -y python3-pip git \ | ||
| build-essential python3-dev libatlas-base-dev |
There was a problem hiding this comment.
Do we need those packages?
Since we only want to download a few python packages, they don't look necessary.
iris_pipeline_project/Dockerfile
Outdated
| RUN apt update && apt install -y python3-pip git \ | ||
| build-essential python3-dev libatlas-base-dev | ||
|
|
||
| RUN pip3 install --break-system-packages numpy pandas scikit-learn |
There was a problem hiding this comment.
Why using --break-system-packages?
iris_pipeline_project/Dockerfile
Outdated
| COPY pipeline/graph.pbtxt /models/iris_pipeline/ | ||
| COPY pipeline/ovmsmodel.py /models/iris_pipeline/ | ||
| COPY model/model.pkl /models/iris_pipeline/ | ||
| COPY model_config.json /model_config.json |
There was a problem hiding this comment.
Let's drop those lines. Eventually we would like to have a Dockerfile for multiple models, datasets etc. so users would build the image with all required packages and mount their data while launching the container.
iris_pipeline_project/convert.py
Outdated
| @@ -0,0 +1,24 @@ | |||
| from sklearn.datasets import load_iris | |||
There was a problem hiding this comment.
This entire file should be wrapped in OvmsPythonModel with training done in execute method
| { | ||
| "custom_node_library_config_list": [], | ||
| "graph_config_list": [ | ||
| { | ||
| "name": "iris_pipeline", | ||
| "base_path": "/models/iris_pipeline", | ||
| "graph_path": "graph.pbtxt" | ||
| } | ||
| ] | ||
| } |
There was a problem hiding this comment.
This format is not correct. See:
https://github.com/openvinotoolkit/model_server/blob/main/docs/python_support/quickstart.md#step-4-prepare-server-configuration-file
for reference
iris_pipeline_project/ovmsmodel.py
Outdated
| @@ -0,0 +1,29 @@ | |||
| import pandas as pd | |||
There was a problem hiding this comment.
This file is unnecessary. After exporting model to ONNX we can use native KServe API support.
| input_stream: "inference_input" | ||
| output_stream: "inference_output" | ||
|
|
||
| node { | ||
| calculator: "InferenceCalculator" | ||
| input_stream: "inference_input" | ||
| output_stream: "inference_output" | ||
| options: { | ||
| [type.googleapis.com/openvino.CalculatorOptions] { | ||
| model_name: "iris_pipeline" | ||
| signature_name: "serving_default" | ||
| } | ||
| } | ||
| } |
There was a problem hiding this comment.
This will not work. See:
https://github.com/openvinotoolkit/model_server/blob/main/docs/python_support/quickstart.md#step-3-prepare-graph-configuration-file
for reference
mzegla
added
the
GSoC
Logistic Regression model trained on iris dataset