hdmjdp opened a new issue #15685: about hybrid_forward URL: https://github.com/apache/incubator-mxnet/issues/15685 class ...(nn.HybridBlock): with self.name_scope(): # 3*3 5*5 7*7 self.weight_dw3 = self.params.get('weight_dw3', shape=(self.num_input // 3, 1, 3), init=mx.init.Xavier(), dtype=self.dtype, allow_deferred_init=True) self.weight_dw5 = self.params.get('weight_dw5', shape=(self.num_input // 3, 1, 5), init=mx.init.Xavier(), dtype=self.dtype, allow_deferred_init=True), self.weight_dw7 = self.params.get('weight_dw7', shape=(self.num_input // 3, 1, 7), init=mx.init.Xavier(), dtype=self.dtype, allow_deferred_init=True) self.bias_dw3 = self.params.get('bias_dw3', shape=(self.num_input // 3,), init=mx.init.Zero(), dtype=self.dtype, allow_deferred_init=True), self.bias_dw5 = self.params.get('bias_dw5', shape=(self.num_input // 3,), init=mx.init.Zero(), dtype=self.dtype, allow_deferred_init=True), self.bias_dw7 = self.params.get('bias_dw7', shape=(self.num_input // 3,), init=mx.init.Zero(), dtype=self.dtype, allow_deferred_init=True) self.g_dw3 = self.params.get('gain_dw3', shape=(self.num_input // 3, 1, 1), init=mx.init.Constant( mx.init.random.uniform(1, np.sqrt(5), shape=(self.num_input // 3, 1, 1))), dtype=self.dtype, allow_deferred_init=True), self.g_dw5 = self.params.get('gain_dw5', shape=(self.num_input // 3, 1, 1), init=mx.init.Constant( mx.init.random.uniform(1, np.sqrt(5), shape=(self.num_input // 3, 1, 1))), dtype=self.dtype, allow_deferred_init=True), self.g_dw7 = self.params.get('gain_dw7', shape=(self.num_input // 3, 1, 1), init=mx.init.Constant( mx.init.random.uniform(1, np.sqrt(5), shape=(self.num_input // 3, 1, 1))), dtype=self.dtype, allow_deferred_init=True) self.weight_pw = self.params.get('weight_pw', shape=(self.num_hidden, self.num_input, 1), init=mx.init.Xavier(), dtype=self.dtype, allow_deferred_init=True) self.bias_pw = self.params.get('bias_pw', shape=(self.num_hidden,), init=mx.init.Zero(), dtype=self.dtype, allow_deferred_init=True) self.g_pw = self.params.get('gain_pw', shape=(self.num_hidden, 1, 1), init=mx.init.Constant( mx.init.random.uniform(1, np.sqrt(5), shape=(self.num_hidden, 1, 1))), dtype=self.dtype, allow_deferred_init=True) self.dropout = nn.Dropout(rate=dropout_rate) def hybrid_forward(self, F, inputs, weight_dw3, weight_dw5, weight_dw7, bias_dw3, bias_dw5, bias_dw7, g_dw3, g_dw5, g_dw7, weight_pw, bias_pw, g_pw, *args, **kwargs): ''' Args: inputs: A 3-D tensor with shape of [batch, depth, time]. Returns: A tensor of the same shape and dtypes as `inputs`. ''' when I call it : self.conv_blocks_textenc[j](tensor), it will give this error: wld, vuv_logits = Vec2Cmp(vec) File "/home/hdm/.local/lib/python3.5/site-packages/mxnet/gluon/block.py", line 541, in __call__ out = self.forward(*args) File "/home/hdm/.local/lib/python3.5/site-packages/mxnet/gluon/block.py", line 918, in forward return self.hybrid_forward(ndarray, x, *args, **params) File "/home/hdm/Documents/dctts/vec2cmp-mx-float/networks_3.py", line 94, in hybrid_forward tensor = self.conv_blocks_textenc[j](tensor) File "/home/hdm/.local/lib/python3.5/site-packages/mxnet/gluon/block.py", line 541, in __call__ out = self.forward(*args) File "/home/hdm/.local/lib/python3.5/site-packages/mxnet/gluon/block.py", line 918, in forward return self.hybrid_forward(ndarray, x, *args, **params) TypeError: hybrid_forward() missing 5 required positional arguments: 'weight_dw5', 'bias_dw3', 'bias_dw5', 'g_dw3', and 'g_dw5'
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services
