cuda.yml

# Copyright 2026 FlagOS Contributors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

# CUDA Hardware Configuration
# This file defines CI/CD settings for CUDA-based testing
# Test configurations are defined in tests/test_utils/config/platforms/cuda.yaml

hardware_name: cuda
display_name: "CUDA Tests"

# Docker image for this hardware
ci_image: harbor.baai.ac.cn/flagos/flagscale:cuda12.8.1-cudnn9.7.1-python3.12-torch2.7.0-time2507111538
ci_train_image: harbor.baai.ac.cn/flagscale/flagscale-train:dev-cu128-py3.12-20260228210721
ci_inference_image: harbor.baai.ac.cn/flagscale/flagscale-inference:dev-cu128-py3.12-20260302102033

# Directory to store image tar files (shared between build and push workflows)
tar_dir: /mnt/airs-business/cicd/image_tar

# Runner labels for this hardware
runner_labels: ['flagscale-nvidia-a100-gpu2-32c-128g']

# Container volumes (hardware-specific paths)
container_volumes:
  - /mnt/airs-business/cicd/baai_datasets:/home/gitlab-runner/data
  - /mnt/airs-business/cicd/baai_tokenizers:/home/gitlab-runner/tokenizers
  - /etc/hosts:/etc/hosts:ro

# Container options (hardware-specific settings)
container_options: >-
  --gpus all
  --shm-size=500g
  --hostname flagscale_cicd
  --user root
  --ulimit nofile=65535:65535
  --cap-add SYS_ADMIN
  --cap-add SYS_PTRACE
  --cap-add IPC_LOCK
  --cap-add NET_ADMIN
  --security-opt seccomp=unconfined

# =============================================================================
# Package Manager Configuration
# =============================================================================
# Supported package managers: pip, uv, conda
#   - pip:   Use pip directly (standard Python)
#   - uv:    Use uv pip (fast, modern package manager)
#   - conda: Use conda environment with pip for PyPI packages
#
# Unified environment parameters:
#   - env_name: Conda environment name (for conda only)
#   - env_path: Environment path (venv path for uv, conda installation path for conda)
#
# To transition to uv in the future:
#   1. Change pkg_mgr to "uv"
#   2. Ensure uv is installed in the Docker image
#   3. Set env_path to the virtual environment path (e.g., "/opt/venv")
#
pkg_mgr: "conda"  # Current: conda for CI/CD compatibility

# Environment path (venv path for uv, conda installation path for conda)
env_path: "/root/miniconda3"

# Conda environment name (for conda only)
env_names:
  train: "flagscale-train"
  hetero_train: "flagscale-train"
  inference: "flagscale-inference"
  rl: "flagscale-rl"