def test_conv3d(self):
module = nnq.Conv3d(3,
3,
kernel_size=3,
stride=1,
padding=0,
dilation=1,
groups=1,
bias=True,
padding_mode="zeros")
self._test_op(module, input_size=[1, 3, 6, 6, 6], generate=False)
def test_conv3d_api(
self, batch_size, in_channels_per_group, D, H, W,
out_channels_per_group, groups, kernel_d, kernel_h, kernel_w,
stride_d, stride_h, stride_w, pad_d, pad_h, pad_w, dilation, X_scale,
X_zero_point, W_scale, W_zero_point, Y_scale, Y_zero_point, use_bias,
use_channelwise, use_fused, qengine,
):
# Tests the correctness of the conv3d module.
if qengine not in torch.backends.quantized.supported_engines:
return
in_channels = in_channels_per_group * groups
out_channels = out_channels_per_group * groups
input_feature_map_size = (D, H, W)
kernel_size = (kernel_d, kernel_h, kernel_w)
stride = (stride_d, stride_h, stride_w)
padding = (pad_d, pad_h, pad_w)
dilation = (dilation, dilation, dilation)
with override_quantized_engine(qengine):
if use_fused:
module_name = "QuantizedConvReLU3d"
qconv_module = nnq_fused.ConvReLU3d(
in_channels, out_channels, kernel_size, stride, padding,
dilation, groups, use_bias, padding_mode="zeros")
else:
module_name = "QuantizedConv3d"
qconv_module = nnq.Conv3d(
in_channels, out_channels, kernel_size, stride, padding,
dilation, groups, use_bias, padding_mode="zeros")
conv_module = nn.Conv3d(
in_channels, out_channels, kernel_size, stride, padding,
dilation, groups, use_bias, padding_mode="zeros")
if use_fused:
relu_module = nn.ReLU()
conv_module = nni.ConvReLU3d(conv_module, relu_module)
conv_module = conv_module.float()
self._test_conv_api_impl(
module_name, qconv_module, conv_module, batch_size,
in_channels_per_group, input_feature_map_size,
out_channels_per_group, groups, kernel_size, stride, padding,
dilation, X_scale, X_zero_point, W_scale, W_zero_point, Y_scale,
Y_zero_point, use_bias, use_fused, use_channelwise)
def test_conv3d(self):
if 'fbgemm' in supported_qengines:
with override_quantized_engine('fbgemm'):
module = nnq.Conv3d(3, 3, kernel_size=3, stride=1, padding=0, dilation=1,
groups=1, bias=True, padding_mode="zeros")
self._test_op(module, input_size=[1, 3, 6, 6, 6], generate=False)
