fix: switch to cu126 index, move torch to [object] extra, update CUDA 12.6 base

This commit is contained in:
Holden
2026-06-10 22:41:48 -04:00
parent cdf10452fb
commit b16488b771
3 changed files with 1204 additions and 1141 deletions
+19 -23
View File
@@ -1,63 +1,59 @@
# ============================================================================= # ── Platform-conditional base image ──────────────────────────────────────
# Stage 1: Base — platform-specific base image # amd64: NVIDIA CUDA 12.6 runtime (matches cu126 torch wheels)
# ============================================================================= # arm64: Plain Ubuntu (CPU-only; no CUDA on ARM64)
# amd64: NVIDIA CUDA runtime (for GPU acceleration)
# arm64: Plain Ubuntu (no CUDA on ARM; CPU or media-kit only)
FROM --platform=$BUILDPLATFORM nvidia/cuda:12.4.1-runtime-ubuntu22.04 AS base-amd64 FROM --platform=$BUILDPLATFORM nvidia/cuda:12.6.3-runtime-ubuntu22.04 AS base-amd64
FROM --platform=$BUILDPLATFORM ubuntu:22.04 AS base-arm64 FROM --platform=$BUILDPLATFORM ubuntu:22.04 AS base-arm64
# ============================================================================= # ── Build stage ───────────────────────────────────────────────────────────
# Stage 2: Build — install system packages + Python + uv
# =============================================================================
ARG TARGETARCH ARG TARGETARCH
FROM base-${TARGETARCH} AS build FROM base-${TARGETARCH} AS build
ENV DEBIAN_FRONTEND=noninteractive ENV DEBIAN_FRONTEND=noninteractive
# Certificates + curl
RUN apt-get update && apt-get install -y --no-install-recommends \ RUN apt-get update && apt-get install -y --no-install-recommends \
ca-certificates \ ca-certificates \
curl \ curl \
gnupg \ gnupg \
&& rm -rf /var/lib/apt/lists/* && rm -rf /var/lib/apt/lists/*
# Python 3.12 + system libs
RUN apt-get update && apt-get install -y --no-install-recommends \ RUN apt-get update && apt-get install -y --no-install-recommends \
software-properties-common \ software-properties-common \
&& add-apt-repository ppa:deadsnakes/ppa -y \ && add-apt-repository ppa:deadsnakes/ppa -y \
&& apt-get update && apt-get install -y --no-install-recommends \ && apt-get update && apt-get install -y --no-install-recommends \
python3.12 python3.12-venv python3.12-dev \ python3.12 python3.12-venv python3.12-dev \
libgl1 libglib2.0-0 libxext6 git g++ tini \ libgl1 libglib2.0-0 libxext6 g++ tini \
&& rm -rf /var/lib/apt/lists/* \ && rm -rf /var/lib/apt/lists/* \
&& ln -sf /usr/bin/python3.12 /usr/bin/python3 && ln -sf /usr/bin/python3.12 /usr/bin/python3
# Install uv
RUN curl -LsSf https://astral.sh/uv/install.sh | sh \ RUN curl -LsSf https://astral.sh/uv/install.sh | sh \
&& cp /root/.local/bin/uv /usr/local/bin/uv && cp /root/.local/bin/uv /usr/local/bin/uv
WORKDIR /app WORKDIR /app
# Copy project files first (for layer caching) # Copy project files for deterministic, cached builds
COPY pyproject.toml uv.lock* ./ COPY pyproject.toml uv.lock ./
COPY if_curator/ if_curator/ COPY if_curator/ if_curator/
COPY entrypoint.sh scheduler.py ./ COPY entrypoint.sh scheduler.py ./
# Platform-conditional Python deps: # Install dependencies — uv resolves per-platform via tool.uv.sources markers
# amd64: install GPU extras (onnxruntime-gpu + CUDA torch) # amd64: --extra gpu --extra object → CUDA torch + onnxruntime-gpu
# arm64: install CPU-only (onnxruntime + torch-cpu) # arm64: --extra object → CPU torch + onnxruntime
RUN if [ "$(uname -m)" = "x86_64" ]; then \ RUN if [ "$TARGETARCH" = "amd64" ]; then \
uv sync --extra gpu --extra object && uv add croniter; \ uv sync --extra gpu --extra object; \
else \ else \
uv sync --extra object --no-binary torch && \ uv sync --extra object; \
UV_TORCH_INDEX_URL=https://download.pytorch.org/whl/cpu uv pip install torch && \
uv add croniter; \
fi \ fi \
&& uv add croniter \
&& uv cache clean && uv cache clean
RUN chmod +x /app/entrypoint.sh RUN chmod +x /app/entrypoint.sh
# ============================================================================= # ── Runtime stage ─────────────────────────────────────────────────────────
# Stage 3: Runtime — minimal image with appuser
# =============================================================================
FROM build AS runtime FROM build AS runtime
RUN groupadd -g 568 apps && useradd -u 568 -g apps -m -s /bin/bash appuser \ RUN groupadd -g 568 apps && useradd -u 568 -g apps -m -s /bin/bash appuser \
+23 -1
View File
@@ -36,11 +36,32 @@ gpu = [
"onnxruntime-gpu>=1.23.2", "onnxruntime-gpu>=1.23.2",
] ]
object = [ object = [
"torch>=2.9.1", "torch>=2.6.0",
"transformers>=4.57.6", "transformers>=4.57.6",
"ultralytics>=8.3.252", "ultralytics>=8.3.252",
] ]
# ── Per-platform PyTorch index configuration ──
# linux/amd64 → CUDA 12.6 wheels (latest PyTorch)
# linux/arm64 → CPU-only wheels
# other → CPU-only wheels
[tool.uv.sources]
torch = [
{ index = "pytorch-cu126", marker = "sys_platform == 'linux' and platform_machine == 'x86_64'" },
{ index = "pytorch-cpu", marker = "sys_platform == 'linux' and platform_machine == 'aarch64'" },
{ index = "pytorch-cpu", marker = "sys_platform != 'linux'" },
]
[[tool.uv.index]]
name = "pytorch-cu126"
url = "https://download.pytorch.org/whl/cu126"
explicit = true
[[tool.uv.index]]
name = "pytorch-cpu"
url = "https://download.pytorch.org/whl/cpu"
explicit = true
[dependency-groups] [dependency-groups]
dev = [ dev = [
"pytest>=8.0", "pytest>=8.0",
@@ -71,6 +92,7 @@ transformers = "transformers"
ultralytics = "ultralytics" ultralytics = "ultralytics"
[tool.deptry.per_rule_ignores] [tool.deptry.per_rule_ignores]
# onnxruntime-gpu is an optional dep that replaces onnxruntime at runtime
DEP002 = ["onnxruntime-gpu"] DEP002 = ["onnxruntime-gpu"]
[build-system] [build-system]
Generated
+1162 -1117
View File
File diff suppressed because it is too large Load Diff