BobbyXue666

MegEngine 1.12.1+cu114 megfile 2.0.0.post2 最小复现如下： ``` class TestModel(M.Module): def __init__(self,): super().__init__() self.conv1 = M.Conv2d(2,16,kernel_size=3, stride=2, padding=1, bias=False) self.bn1 = M.BatchNorm2d(16) self.relu1 = M.ReLU() self.conv2 = M.Conv2d(16,4,kernel_size=3, stride=2, padding=1, bias=False) self.bn2...

这是我的json文件 ``` { "dump_dir":"./batch_dump/", "models":[ { "model_name":"test", "model_path":"./example/test.mge", "input_shape_str":"data=(1,2,20,20)", "enable_nchw44":true } ], "cv":{ "transpose":["ui8"], "roicopy":["ui8"], "rotate":["ui8"], "flip":["ui8"], "resize_linear":["ui8"], "warp_affine_replicate_linear":["ui8"], "rgb2bgr":["ui8"], "yuv2bgr_nv21":["ui8"], "rgb2yuv":["ui8"] } } ``` 这是生成模型的命令 `./script/ppl_gen.sh ./bin/mgb-to-tinynn ./example/demo.json test...

将bias去掉可以编过，想知道是为什么 `self.conv1 = M.Conv2d(2,16,kernel_size=3, stride=2, padding=1)`