Tensor Manipulation: squeeze(), detach(), cpu() in PyTorch

This\u0020line\u0020of\u0020code\u0020performs\u0020three\u0020operations\u0020on\u0020the\u0020tensor\u0020output_tensor:\n\n1.\u0020squeeze(): This\u0020function\u0020removes\u0020all\u0020the\u0020dimensions\u0020of\u0020size\u00201\u0020from\u0020the\u0020tensor.\u0020It\u0020reduces\u0020the\u0020dimensions\u0020of\u0020the\u0020tensor\u0020and\u0020returns\u0020a\u0020new\u0020tensor\u0020with\u0020the\u0020squeezed\u0020dimensions.\u0020This\u0020is\u0020useful\u0020when\u0020working\u0020with\u0020tensors\u0020that\u0020have\u0020unnecessary\u0020dimensions.\n\n2.\u0020detach(): This\u0020function\u0020detaches\u0020the\u0020tensor\u0020from\u0020the\u0020computation\u0020graph.\u0020It\u0020returns\u0020a\u0020new\u0020tensor\u0020that\u0020shares\u0020the\u0020same\u0020data\u0020but\u0020is\u0020not\u0020connected\u0020to\u0020the\u0020computation\u0020history.\u0020This\u0020is\u0020useful\u0020when\u0020you\u0020want\u0020to\u0020stop\u0020further\u0020gradient\u0020calculations\u0020on\u0020a\u0020tensor.\n\n3.\u0020cpu(): This\u0020function\u0020moves\u0020the\u0020tensor\u0020from\u0020any\u0020device\u0020(such\u0020as\u0020GPU)\u0020to\u0020the\u0020CPU.\u0020It\u0020returns\u0020a\u0020new\u0020tensor\u0020that\u0020resides\u0020in\u0020the\u0020CPU\u0020memory.\u0020This\u0020is\u0020useful\u0020when\u0020you\u0020want\u0020to\u0020perform\u0020operations\u0020on\u0020the\u0020tensor\u0020using\u0020a\u0020CPU.\n\nThe\u0020result\u0020of\u0020these\u0020operations\u0020is\u0020assigned\u0020to\u0020a\u0020new\u0020tensor\u0020called\u0020output_tensor.