[Feature] Partial steps in batched envs #5772

	name: RLHF Tests on Linux

	on:
	pull_request:
	push:
	branches:
	- nightly
	- main
	- release/*
	workflow_dispatch:

	concurrency:
	# Documentation suggests ${{ github.head_ref }}, but that's only available on pull_request/pull_request_target triggers, so using ${{ github.ref }}.
	# On master, we want all builds to complete even if merging happens faster to make it easier to discover at which point something broke.
	group: ${{ github.workflow }}-${{ github.ref == 'refs/heads/main' && format('ci-master-{0}', github.sha) \|\| format('ci-{0}', github.ref) }}
	cancel-in-progress: true

	jobs:
	unittests:
	strategy:
	matrix:
	python_version: ["3.9"]
	cuda_arch_version: ["12.1"]
	uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
	with:
	repository: pytorch/rl
	runner: "linux.g5.4xlarge.nvidia.gpu"
	# gpu-arch-type: cuda
	# gpu-arch-version: "11.7"
	docker-image: "nvidia/cudagl:11.4.0-base"
	timeout: 120
	script: \|
	if [[ "${{ github.ref }}" =~ release/* ]]; then
	export RELEASE=1
	export TORCH_VERSION=stable
	else
	export RELEASE=0
	export TORCH_VERSION=nightly
	fi

	set -euo pipefail
	export PYTHON_VERSION="3.9"
	export CU_VERSION="cu117"
	export TAR_OPTIONS="--no-same-owner"
	export UPLOAD_CHANNEL="nightly"
	export TF_CPP_MIN_LOG_LEVEL=0
	export TD_GET_DEFAULTS_TO_NONE=1

	bash .github/unittest/linux_libs/scripts_rlhf/setup_env.sh
	bash .github/unittest/linux_libs/scripts_rlhf/install.sh
	bash .github/unittest/linux_libs/scripts_rlhf/run_test.sh
	bash .github/unittest/linux_libs/scripts_rlhf/post_process.sh

Provide feedback