Extracted workload for stride=2 is off by 1
{"v": 0.1, "r": [[0.0004028920402144772, 0.0004029741474530831, 0.00040302522520107236], 0, 1.6767456531524658, 1555614243.9415061], "i": ["cuda -model=unknown", "topi_nn_conv2d", [["TENSOR", [1, 3, 1201, 1921], "float32"], ["TENSOR", [32, 3, 3, 3], "float32"], [2, 2], [0, 0], [1, 1], "NCHW", "float32"], {}, ["conv2d", [1, 3, 1201, 1921, "float32"], [32, 3, 3, 3, "float32"], [2, 2], [0, 0], [1, 1], "NCHW", "float32"], {"e": [["tile_f", "sp", [1, 2, 16, 1]], ["tile_y", "sp", [600, 1, 1, 1]], ["tile_x", "sp", [5, 12, 16, 1]], ["tile_rc", "sp", [3, 1]], ["tile_ry", "sp", [1, 3]], ["tile_rx", "sp", [3, 1]], ["auto_unroll_max_step", "ot", 512], ["unroll_explicit", "ot", 0]], "t": "direct", "c": null, "i": 609459219}]}```