feat(install_cuda): add CI mode with automatic disk cleanup for GitHub Actions

Add CI_MODE environment variable to enable CI-specific optimizations:

## New Features
- CI_MODE=1 enables automatic disk cleanup before installation
- Removes common pre-installed software on GitHub Actions runners:
  - .NET SDK, Android SDK, GHC, CodeQL, Boost, Swift, GraalVM, etc.
- Cleans apt cache and Docker images when available
- Displays disk space statistics before and after cleanup:
  - Total disk size, used space, available space, usage percentage
  - Calculates and shows freed disk space

## New Environment Variables
- CI_MODE: Set to 1 to enable CI optimizations (default: 0)
- REQUIRED_DISK_SPACE_GB: Minimum required disk space in GB (default: 15)

## Usage
```bash
# GitHub Actions / CI environment
CI_MODE=1 CUDA_INSTALL_PREFIX=/tmp/cuda ./install_cuda.sh 12.8

# Local installation (unchanged behavior)
CUDA_INSTALL_PREFIX=~/opt ./install_cuda.sh 12.8
```

This should resolve 'No space left on device' errors on GitHub Actions runners
by freeing approximately 30-50GB of disk space before CUDA installation.

Author: FindHao

install_cuda.sh

@@ -1,5 +1,6 @@
 #!/bin/bash
 # Usage: CUDA_INSTALL_PREFIX=/home/yhao/opt ./install_cuda.sh 12.8
+# CI Usage: CI_MODE=1 CUDA_INSTALL_PREFIX=/tmp/cuda ./install_cuda.sh 12.8
 # Notice: Part of this script is synced with https://github.com/pytorch/pytorch/blob/main/.ci/docker/common/install_cuda.sh
 set -ex
 
@@ -23,9 +24,16 @@ CUDA_VERSION=${CUDA_VERSION:-12.8}
 NVSHMEM_VERSION=${NVSHMEM_VERSION:-3.4.5}
 INSTALL_NCCL=${INSTALL_NCCL:-1}
 
+# CI mode settings
+# Set CI_MODE=1 to enable CI-specific optimizations (disk cleanup, etc.)
+CI_MODE=${CI_MODE:-0}
+# Minimum required disk space in GB (can be overridden)
+REQUIRED_DISK_SPACE_GB=${REQUIRED_DISK_SPACE_GB:-15}
+
 echo "CUDA_INSTALL_PREFIX=${CUDA_INSTALL_PREFIX}"
 echo "CUDA_VERSION=${CUDA_VERSION}"
 echo "INSTALL_NCCL=${INSTALL_NCCL}"
+echo "CI_MODE=${CI_MODE}"
 
 # Version configuration using associative arrays
 declare -A CUDA_FULL_VERSION=(
@@ -179,6 +187,103 @@ function check_network {
   fi
 }
 
+# Get disk space information for a given path
+# Returns: total_gb used_gb available_gb use_percent
+function get_disk_info {
+  local target_path="${1:-/}"
+  df -BG "${target_path}" | awk 'NR==2 {
+    gsub(/G/, "", $2); gsub(/G/, "", $3); gsub(/G/, "", $4); gsub(/%/, "", $5);
+    print $2, $3, $4, $5
+  }'
+}
+
+# Print disk space summary
+function print_disk_summary {
+  local label="$1"
+  local target_path="${2:-/}"
+
+  local disk_info
+  disk_info=$(get_disk_info "${target_path}")
+  local total_gb=$(echo "${disk_info}" | awk '{print $1}')
+  local used_gb=$(echo "${disk_info}" | awk '{print $2}')
+  local available_gb=$(echo "${disk_info}" | awk '{print $3}')
+  local use_percent=$(echo "${disk_info}" | awk '{print $4}')
+
+  echo "💾 ${label}:"
+  echo "   Total: ${total_gb}GB | Used: ${used_gb}GB | Available: ${available_gb}GB | Usage: ${use_percent}%"
+}
+
+# CI mode: Free up disk space by removing unnecessary pre-installed software
+# This is useful for GitHub Actions runners which have limited disk space
+function ci_free_disk_space {
+  if [ "${CI_MODE}" != "1" ]; then
+    echo "CI_MODE not enabled, skipping disk cleanup"
+    return 0
+  fi
+
+  echo "🧹 ===== CI Mode: Freeing up disk space ====="
+
+  # Print disk space before cleanup
+  print_disk_summary "Disk space BEFORE cleanup" "/"
+  local before_available
+  before_available=$(get_disk_info "/" | awk '{print $3}')
+
+  echo ""
+  echo "Removing unnecessary software packages..."
+
+  # List of directories to remove (common on GitHub Actions runners)
+  local dirs_to_remove=(
+    "/usr/share/dotnet"
+    "/usr/local/lib/android"
+    "/opt/ghc"
+    "/opt/hostedtoolcache/CodeQL"
+    "/usr/local/share/boost"
+    "/usr/share/swift"
+    "/usr/local/graalvm"
+    "/usr/local/.ghcup"
+    "/opt/hostedtoolcache/Python"
+    "/opt/hostedtoolcache/Ruby"
+    "/opt/hostedtoolcache/go"
+    "/opt/hostedtoolcache/node"
+  )
+
+  local removed_count=0
+  for dir in "${dirs_to_remove[@]}"; do
+    if [ -d "${dir}" ]; then
+      echo "  Removing ${dir}..."
+      sudo rm -rf "${dir}" 2>/dev/null || rm -rf "${dir}" 2>/dev/null || true
+      ((removed_count++)) || true
+    fi
+  done
+
+  echo "  Removed ${removed_count} directories"
+
+  # Clean apt cache if apt is available
+  if command_exists apt-get; then
+    echo "  Cleaning apt cache..."
+    sudo apt-get clean 2>/dev/null || true
+    sudo apt-get autoremove -y 2>/dev/null || true
+  fi
+
+  # Clean docker if available
+  if command_exists docker; then
+    echo "  Cleaning Docker images..."
+    docker system prune -af 2>/dev/null || true
+  fi
+
+  echo ""
+  # Print disk space after cleanup
+  print_disk_summary "Disk space AFTER cleanup" "/"
+  local after_available
+  after_available=$(get_disk_info "/" | awk '{print $3}')
+
+  # Calculate freed space
+  local freed_gb=$((after_available - before_available))
+  echo ""
+  echo "✅ Freed approximately ${freed_gb}GB of disk space"
+  echo ""
+}
+
 # Check if file or directory exists
 function check_exists {
   if [ ! -e "$1" ]; then
@@ -192,8 +297,11 @@ check_exists "${CUDA_INSTALL_PREFIX}"
 # Check dependencies
 check_dependencies
 
+# CI mode: Free up disk space before checking disk space requirements
+ci_free_disk_space
+
 # Check disk space
-check_disk_space 15
+check_disk_space "${REQUIRED_DISK_SPACE_GB}"
 
 # Check network connectivity
 check_network