[C++] Invoke storage allocation for CUDA Graph explicitly #1899

	# GH actions.
	# We use it to cover windows builds
	# Jenkins is still the primary CI
	name: Windows CI

	on:
	push:
	branches:
	- main
	pull_request:
	branches:
	- main

	jobs:
	Windows:
	runs-on: windows-latest
	defaults:
	run:
	shell: 'cmd /C call {0}'

	steps:
	- name: Git config
	run: >-
	git config --system core.longpaths true
	- uses: actions/checkout@v3
	with:
	submodules: 'recursive'
	- uses: conda-incubator/setup-miniconda@v2
	with:
	activate-environment: mlc-llm-build
	channel-priority: strict
	environment-file: ci/build-environment.yaml
	auto-activate-base: false
	- name: Conda info
	run: \|
	conda info
	conda list
	python --version
	- name: Build MLC-LLM
	run: >-
	ci/task/build_win.bat

Provide feedback