conv_tbc(i, w, b, pad), (inp, weight, bias, 3)) def run_conv_double_back_test(self, kern, stride, padding, chan_in, chan_out, batch_size, inp_size, dilation, no_weight, groups=1, use_cuda=False, use_bias=True): tensor = torch.Tensor(1) 3 changes: 3 additions & 0 ...
Testing: https://github.com/facebookresearch/fairseq-py/blob/master/tests/test_convtbc.py Notes: Temporary exception in tools/autograd/gen_variable_type.py that we need to address later. Follow-up: New PR to https://github.com/facebookresearch/fairseq-py to switch over to this implementation...