Move unittest invokation to _ci due to workflow nested call limit #3

	name: ~test TransformerEngine

	on:
	workflow_call:
	inputs:
	TE_IMAGE:
	type: string
	description: 'JAX+TE+PAXML image'
	required: true
	default: 'ghcr.io/nvidia/upstream-pax:latest'
	ARTIFACT_PREFIX:
	type: string
	description: 'Name of the artifact zip file'
	required: false
	default: 'te'

	jobs:
	# te-multi-gpu:
	# uses: ./.github/workflows/_test_slurm_pyxis.yaml
	# strategy:
	# matrix:
	# N_GPU: [2, 4, 8]
	# fail-fast: false
	# secrets:
	# SSH_PRIVATE_KEY: ${{ secrets.SSH_PRIVATE_KEY }}
	# SLURM_LOGIN_USER: ${{ secrets.CLUSTER_LOGIN_USER }}
	# CONTAINER_REGISTRY_TOKEN: ${{ secrets.github_token }}
	# with:
	# NAME: ${{ inputs.ARTIFACT_PREFIX }}-${{ matrix.N_GPU }}GPU
	# SLURM_LOGIN_HOSTNAME: ${{ vars.HOSTNAME_SLURM_LOGIN }}
	# OUTPUT_BASEDIR: /nfs/cluster
	# OUTPUT_MOUNTPOINT: /output
	# NODES: 1
	# GPUS_PER_NODE: ${{ matrix.N_GPU }}
	# NTASKS: 1
	# NTASKS_PER_NODE: 1
	# TIME_LIMIT: '00:10:00'
	# EXTRA_EXPORTS: 'VOCAB_PATH=gs://t5-data/vocabs/cc_all.32000.100extra/sentencepiece.model'
	# IMAGE: ${{ inputs.TE_IMAGE }}
	# SRUN_PREAMBLE: \|
	# nvidia-smi
	# pip install \
	# pytest \
	# pytest-reportlog \
	# cuda-python \
	# -r ${SRC_PATH_TRANSFORMER_ENGINE}/examples/jax/encoder/requirements.txt
	# SRUN_SCRIPT: \|
	# set -ex
	# cd ${SRC_PATH_TRANSFORMER_ENGINE}/examples/jax/encoder
	# pytest --report-log=/output/pytest-report.jsonl \
	# test_single_gpu_encoder.py \
	# test_multigpu_encoder.py \
	# test_model_parallel_encoder.py

Provide feedback