PyTorch with Clutser or MicroCloud

Hello,
I have three laptops with Ubuntu on all of them. Configuration as follows:
Laptop1 : i5 processor with Intel GPU 16GB RAM
Laptop2: i3 processor with Intel GPU 12GB RAM
Laptop3: i7 processor with Intel GPU 6GB RAM.

** Now I want to pool the resources viz. CPU+GPU+RAM ** Not so keen on pooling the secondary memory i.e storage space.

Dear Members, Please let me know what do you suggest Cluster or Micro Cloud. Which option suites PyTorch. I am not keen on Spark or Yarn solution using Pyspark. Later I intend to add Jetson nano graphics card to this cluster or cloud whichever you suggest. Any suggestions in this regard would be highly appreciated. Already I am seeing 1.6X improvement in speed with PyTorch for Convolution Neural Network with MNIST data set visa vis TensorFlow without Mother board getting heated on single node. Back bone for the Nodes is Wireless Local Area Network with practical speed of 5 Mega B(b)its per second. Any suggestions in this regard would be highly appreciated. Thanks in advance