Upload Dockerfile with huggingface_hub

Dockerfile

@@ -1,166 +1,152 @@
+# syntax=docker/dockerfile:1.7-labs
+
+# ─────────────────────────────────────────────────────────────
+# IMGFLOW — Production Dockerfile
+# Target: HuggingFace Spaces (CPU, x86_64, port 7860)
+#
+# Base image rationale — python:3.11-slim-bookworm:
+#   • Alpine/musl: INCOMPATIBLE — torch, numpy, scipy, onnxruntime
+#     all ship manylinux (glibc) wheels only; musl libc breaks them
+#   • distroless: no shell → can't run healthcheck curl or venv activate
+#   • slim-bookworm: smallest Debian 12 (glibc 2.36) image that runs
+#     the full stack; libstdc++6 pre-installed; verified compatible
+#     with every wheel in requirements.txt
+#   • bullseye (Debian 11): glibc 2.31 also works but older security
+#     patches and HF build runners default to bookworm
+# ─────────────────────────────────────────────────────────────
+
 # ============================================================
 # Stage 1 — Builder
-# Ultra-optimized wheel + model builder
 # ============================================================
 FROM python:3.11-slim-bookworm AS builder
 
 ENV DEBIAN_FRONTEND=noninteractive \
     PYTHONDONTWRITEBYTECODE=1 \
     PYTHONUNBUFFERED=1 \
+    # Leave pip cache enabled here — mount cache handles it efficiently
     PIP_DISABLE_PIP_VERSION_CHECK=1 \
-    PIP_NO_CACHE_DIR=1 \
     PIP_ROOT_USER_ACTION=ignore \
     VIRTUAL_ENV=/opt/venv \
-    PATH="/opt/venv/bin:$PATH"
+    PATH="/opt/venv/bin:$PATH" \
+    TORCH_HOME=/opt/models/torch \
+    U2NET_HOME=/opt/models/u2net
 
-# Only absolute minimum build dependencies
+# Build-time deps
+# libGL + libglib2.0-0 are needed by cv2 at import time (rembg pre-download step)
 RUN apt-get update && apt-get install -y --no-install-recommends \
-    gcc \
-    g++ \
-    curl \
-    ca-certificates \
+        gcc \
+        curl \
+        ca-certificates \
+        libgl1 \
+        libglib2.0-0 \
     && rm -rf /var/lib/apt/lists/*
 
-# Create isolated venv
 RUN python -m venv $VIRTUAL_ENV
 
 WORKDIR /app
-
 COPY backend/requirements.txt .
 
-# Faster pip + smaller install
-RUN pip install --upgrade pip wheel setuptools
+RUN pip install --upgrade pip setuptools wheel packaging
 
-# Install CPU-only torch first
+# PyTorch CPU wheel must be installed BEFORE iopaint to avoid torch version
+# conflicts. Explicit CPU index prevents pip pulling a 3 GB CUDA build.
 RUN --mount=type=cache,target=/root/.cache/pip \
-    pip install --no-cache-dir \
-    torch torchvision \
-    --index-url https://download.pytorch.org/whl/cpu
+    pip install torch torchvision --index-url https://download.pytorch.org/whl/cpu
 
-# Install remaining deps
 RUN --mount=type=cache,target=/root/.cache/pip \
-    pip install --no-cache-dir -r requirements.txt
-
-# Copy backend only
-COPY backend/main.py .
-COPY backend/processing.py .
+    pip install -r requirements.txt
 
-# ============================================================
-# Pre-download models
-# ============================================================
+COPY backend/main.py backend/processing.py ./
 
-ENV U2NET_HOME=/opt/models/u2net
+# ── Pre-download AI models at build time (zero cold-start delay) ──────────
 
+# ISNet background removal model (~170 MB)
 RUN python -c "\
 from rembg import new_session; \
-new_session('isnet-general-use')"
+new_session('isnet-general-use'); \
+print('✓ ISNet ready')"
+
+# LaMa inpainting model (~200 MB) — curl avoids importing iopaint.model
+# which would chain-import diffusers/anytext and fail before torch is ready
+RUN mkdir -p /opt/models/torch/hub/checkpoints && \
+    curl -fsSL --retry 3 --retry-delay 2 \
+        -o /opt/models/torch/hub/checkpoints/big-lama.pt \
+        https://github.com/Sanster/models/releases/download/add_big_lama/big-lama.pt && \
+    echo "✓ LaMa ready"
+
+# ── Aggressive venv cleanup (~40–80 MB saved) ─────────────────────────────
+RUN find $VIRTUAL_ENV \( \
+        -type d -name '__pycache__' -o \
+        -type d -name 'tests'       -o \
+        -type d -name 'test'        -o \
+        -type d -name '*.dist-info' -o \
+        -type f -name '*.pyc'       -o \
+        -type f -name '*.pyo'       -o \
+        -type f -name '*.pyd' \
+    \) -exec rm -rf {} + 2>/dev/null || true
 
-RUN python -c "\
-from iopaint.model.lama import LaMa, LAMA_MODEL_URL, LAMA_MODEL_MD5; \
-from iopaint.helper import download_model; \
-download_model(LAMA_MODEL_URL, LAMA_MODEL_MD5)"
 
 # ============================================================
-# Aggressive cleanup
-# ============================================================
-
-RUN find /opt/venv -type d -name '__pycache__' -exec rm -rf {} + && \
-    find /opt/venv -type d -name 'tests' -exec rm -rf {} + && \
-    find /opt/venv -type d -name 'test' -exec rm -rf {} + && \
-    find /opt/venv -type f -name '*.pyc' -delete && \
-    find /opt/venv -type f -name '*.pyo' -delete && \
-    find /opt/venv -type f -name '*.a' -delete && \
-    find /opt/venv -type f -name '*.so.debug' -delete && \
-    strip --strip-unneeded /opt/venv/lib/python3.11/site-packages/**/*.so 2>/dev/null || true
-
-# ============================================================
-# Stage 2 — Runtime
-# Distroless-style slim runtime
+# Stage 2 — Runtime  (inherits nothing from builder except COPY)
 # ============================================================
 FROM python:3.11-slim-bookworm
 
-LABEL maintainer="Ultimate-AI-Container" \
-      description="Extreme optimized FastAPI + rembg + LaMa container"
+LABEL org.opencontainers.image.description="IMGFLOW — FastAPI image processing for HuggingFace Spaces"
 
-# ============================================================
-# Runtime environment optimizations
-# ============================================================
-
-ENV DEBIAN_FRONTEND=noninteractive \
-    PYTHONDONTWRITEBYTECODE=1 \
+ENV PYTHONDONTWRITEBYTECODE=1 \
     PYTHONUNBUFFERED=1 \
     PYTHONFAULTHANDLER=1 \
-    PYTHONHASHSEED=random \
-    PIP_DISABLE_PIP_VERSION_CHECK=1 \
-    PIP_ROOT_USER_ACTION=ignore \
     VIRTUAL_ENV=/opt/venv \
     PATH="/opt/venv/bin:$PATH" \
     PYTHONPATH=/app \
+    TORCH_HOME=/opt/models/torch \
     U2NET_HOME=/opt/models/u2net \
-    HF_HOME=/tmp/huggingface \
-    MPLCONFIGDIR=/tmp/matplotlib \
-    TORCH_HOME=/tmp/torch \
-    OMP_NUM_THREADS=1 \
-    OPENBLAS_NUM_THREADS=1 \
-    MKL_NUM_THREADS=1 \
-    NUMEXPR_NUM_THREADS=1 \
-    VECLIB_MAXIMUM_THREADS=1 \
+    # Thread pinning: prevents CPU oversubscription under single-worker uvicorn
+    OMP_NUM_THREADS=2 \
+    OPENBLAS_NUM_THREADS=2 \
+    MKL_NUM_THREADS=2 \
+    NUMEXPR_NUM_THREADS=2 \
+    VECLIB_MAXIMUM_THREADS=2 \
+    # Limit glibc malloc arena growth (saves ~50–150 MB RSS on long-running AI workloads)
     MALLOC_ARENA_MAX=2 \
     OPENCV_OPENCL_RUNTIME=disabled \
-    OPENCV_IO_ENABLE_OPENEXR=0 \
     ORT_DISABLE_TELEMETRY=1
 
-# Runtime-only libs
+# Runtime native libs — deliberately minimal:
+#   libglib2.0-0  : cv2 links libglib-2.0.so.0 (not bundled in headless wheel)
+#   libgomp1      : OpenMP used by numpy, scipy, torch CPU kernels
+#   curl          : healthcheck only — wget not in slim-bookworm by default
+# NOT needed (already in slim-bookworm or bundled by wheel):
+#   libstdc++6    : pre-installed in every slim image
+#   libgl1        : cv2 >= 4.6 headless bundles its own libGL stub
 RUN apt-get update && apt-get install -y --no-install-recommends \
-    libglib2.0-0 \
-    libgomp1 \
-    libstdc++6 \
-    libgl1 \
-    curl \
-    && rm -rf /var/lib/apt/lists/*
+        libglib2.0-0 \
+        libgomp1 \
+        curl \
+    && apt-get clean && rm -rf /var/lib/apt/lists/*
 
-# Non-root user
-RUN useradd -r -u 1001 -g users appuser
+RUN useradd -m -u 1001 appuser
 
 WORKDIR /app
 
-# Copy only final runtime artifacts
-COPY --from=builder /opt/venv /opt/venv
-COPY --from=builder /opt/models /opt/models
-
-# Backend
-COPY backend/main.py .
-COPY backend/processing.py .
-
-# Static assets
-COPY index.html ./static/index.html
-COPY style.css ./static/style.css
-COPY script.js ./static/script.js
+COPY --from=builder /opt/venv    /opt/venv
+COPY --from=builder /opt/models  /opt/models
+COPY backend/main.py backend/processing.py ./
+COPY index.html style.css script.js ./static/
 RUN wget -O ./static/favicon.ico "https://raw.githubusercontent.com/abir614/IMGFLOW/refs/heads/main/favicon.ico"
 
-# Permissions
-RUN mkdir -p /app/static && \
-    chown -R appuser:users /app /opt/models
+RUN chown -R appuser:appuser /app /opt/models
 
 USER appuser
 
-EXPOSE 7860
-
-# ============================================================
-# Healthcheck
-# ============================================================
-
-HEALTHCHECK --interval=30s --timeout=5s --start-period=20s --retries=3 \
-CMD curl -fsS http://127.0.0.1:7860/api/health || exit 1
+HEALTHCHECK --interval=30s --timeout=5s --start-period=30s --retries=3 \
+    CMD curl -fsS http://127.0.0.1:7860/api/health || exit 1
 
-# ============================================================
-# Launch
-# ============================================================
+EXPOSE 7860
 
 CMD ["uvicorn", "main:app", \
      "--host", "0.0.0.0", \
      "--port", "7860", \
      "--workers", "1", \
      "--loop", "uvloop", \
-     "--http", "httptools", \
-     "--no-access-log"]
+     "--http", "httptools"]