Dfp pclr yoda

[kilonzi]

This pull request introduces a deployment pipeline for the PCLR model, including schema definitions, Docker containerization, and preprocessing/postprocessing scripts. The changes focus on defining input/output formats, creating a Dockerized environment for processing, and implementing scripts for preparing and finalizing data.

Deployment Pipeline Setup:

• Model Schema Definition:
  Added a JSON schema in pclr_model_schema.json to specify the model's input (ecg_tensor) and output (embed) formats, including their shapes and data types.

• Dockerfile for Processing:
  Created a Dockerfile to set up a lightweight Python 3.9-based environment for running the preprocessing (prepare.py) and postprocessing (finalize.py) scripts. It installs dependencies from requirements.txt and sets the entry point to Python.

Data Processing Scripts:

• Preprocessing Script (prepare.py):
  Added a script to process raw ECG files into HDF5 tensor format. It reads input CSVs, extracts ECG data from files, interpolates and normalizes the data, and saves it in a structured HDF5 format.

• Postprocessing Script (finalize.py):
  Added a script to merge model predictions with input metadata. It reads a CSV of metadata and a JSON of predictions, validates dimensions, and outputs a combined CSV with embeddings appended.

• Dependencies:
  Added a requirements.txt file listing the necessary Python libraries (pandas, numpy, h5py, smart-open[gcs]) for preprocessing and postprocessing.