Migrate and optimize ULTIMATE_HYBRID_MODEL (workflow-005) with GPU support

Author: Abdelrahman-Mohamed-Taha-MAHMOUD

apps/q/qsar-hybrid-model/Dockerfile

@@ -0,0 +1,59 @@
+FROM mambaorg/micromamba:latest
+
+# Set working directory
+WORKDIR /app
+
+# Metadata
+LABEL version="3.0"
+LABEL description="QSAR Hybrid Model V3 - Interactive Dashboards & Prediction Pipeline"
+LABEL maintainer="QSAR Abdo"
+
+# Copy environment file
+COPY final_environment.yml /tmp/env.yaml
+
+# Install dependencies
+RUN micromamba install -y -n base -f /tmp/env.yaml && \
+    micromamba clean --all --yes
+
+# Activate environment by default
+ARG MAMBA_DOCKERFILE_ACTIVATE=1
+ENV PATH="/opt/conda/bin:${PATH}"
+ENV LD_LIBRARY_PATH="/opt/conda/lib:/opt/conda/lib/python3.10/site-packages/nvidia/cudnn/lib:/opt/conda/lib/python3.10/site-packages/nvidia/cublas/lib:/opt/conda/lib/python3.10/site-packages/nvidia/cuda_runtime/lib:/opt/conda/lib/python3.10/site-packages/nvidia/cuda_cupti/lib:/opt/conda/lib/python3.10/site-packages/nvidia/cuda_nvcc/lib:/opt/conda/lib/python3.10/site-packages/nvidia/cufft/lib:/opt/conda/lib/python3.10/site-packages/nvidia/curand/lib:/opt/conda/lib/python3.10/site-packages/nvidia/cusolver/lib:/opt/conda/lib/python3.10/site-packages/nvidia/cusparse/lib:/opt/conda/lib/python3.10/site-packages/nvidia/nccl/lib:/opt/conda/lib/python3.10/site-packages/nvidia/nvjitlink/lib"
+
+# TensorFlow & NVIDIA Optimizations
+ENV TF_ENABLE_ONEDNN_OPTS=0
+ENV TF_CPP_MIN_LOG_LEVEL=2
+ENV NVIDIA_VISIBLE_DEVICES=all
+ENV NVIDIA_DRIVER_CAPABILITIES=compute,utility
+
+# Switch to root to perform file operations
+USER root
+
+# Copy source code (respecting .dockerignore if present, but we copy widely here)
+# We copy specific directories to avoid clutter
+COPY 1_data_prep /app/1_data_prep
+COPY 2_model_train /app/2_model_train
+COPY 3_analyze_overfitting /app/3_analyze_overfitting
+COPY 4_predict_from_csv /app/4_predict_from_csv
+COPY input_files /app/input_files
+
+# Create output directories to ensure they exist/persist
+RUN mkdir -p /app/1_data_prep/outputs \
+    /app/2_model_train/outputs \
+    /app/3_analyze_overfitting/outputs \
+    /app/4_predict_from_csv/outputs
+
+# Make run scripts executable
+RUN chmod +x /app/1_data_prep/run.sh \
+    /app/2_model_train/run.sh \
+    /app/3_analyze_overfitting/run.sh \
+    /app/4_predict_from_csv/run.sh
+
+# Switch back to mambauser
+USER $MAMBA_USER
+
+# Expose data volumes for explicit input/output handling
+VOLUME ["/app/input_files", "/app/4_predict_from_csv/inputs"]
+
+# Default command: Open a shell
+CMD ["/bin/bash"]

apps/q/qsar-hybrid-model/final_environment.yml

@@ -0,0 +1,14 @@
+name: qsar_hybrid_env
+channels:
+  - conda-forge
+  - defaults
+dependencies:
+  - python=3.10
+  - numpy
+  - pandas
+  - matplotlib
+  - scikit-learn
+  - rdkit
+  - pip
+  - pip:
+    - tensorflow[and-cuda]

workflows/workflow-005/.chiral/workflow.toml

@@ -1,12 +1,12 @@
 name = "QSAR_Modeling_Pipeline"
-description = "Machine Learning-Based Drug Discovery with Premium Interactive Visualization"
+description = "Machine Learning-Based Drug Discovery with Interactive Visualization"
 
 [dependencies]
 # Node 2 depends on Node 1
-02-feature-engineering = ["01-data-preparation"]
+2_model_train = ["1_data_prep"]
 
-# Node 3 depends on Node 2
-03-model-training = ["02-feature-engineering"]
+# Node 3 depends on Node 1 (data) and Node 2 (model)
+3_analyze_overfitting = ["1_data_prep", "2_model_train"]
 
-# Node 4 depends on Node 2 (scaler, ad_stats) and Node 3 (model)
-04-prediction = ["02-feature-engineering", "03-model-training"]
+# Node 4 depends on Node 2 (model & pipeline)
+4_predict_from_csv = ["2_model_train"]