Am I initializing the hidden state of the RNN correctly? Why is my validation accuracy not improving

Hello, here is the code to my MLP-GRU model:


class MLP(nn.Module):
  
  def __init__(self, num_classes, rnn_layers, hidden_size, fc_size):
    super().__init__()
    self.hidden_size = hidden_size
    self.num_classes = num_classes
    self.fc_size = fc_size
    

    self.fc_pre= nn.Sequential(nn.Linear(201, 120),
                               nn.Dropout(p=0.5),
                           nn.ReLU(),
                           nn.Linear(120, 75),
                           nn.Dropout(p=0.5),
                           nn.ReLU(),
                           nn.Linear(75, fc_size),
                           nn.Dropout(p=0.5),
                           nn.ReLU())
    #Defines the number of features that define each element (time-stamp) of the input sequence
    self.rnn = nn.GRU(input_size = fc_size,
                hidden_size = hidden_size,
                num_layers = rnn_layers,
                batch_first = True)
    self.fc = nn.Linear(hidden_size, num_classes)
    #feed the previous layer output to softmax and do softmax on the class dimension
    #self.softmax = nn.Softmax(dim=-1)
  
  #initialize hidden state for RNN  
  def init_hidden(self, num_layers, batch_size):
         return (torch.zeros(num_layers, batch_size, self.hidden_size).cuda(),
                 torch.zeros(num_layers, batch_size, self.hidden_size).cuda())

  def forward(self, inputs, hidden=None, steps=0):
        length = len(inputs)
        # print('inputs size')
        # print(inputs[3].size())
        #Input data: RNN should have 3 dimensions. (Batch Size, Sequence Length and Input Dimension (the number of expected features which is 201)
        fs = torch.zeros(inputs[0].size(0), length, self.rnn.input_size).cuda()

        for i in range(length):
            f = inputs[i]
           
            #flattens the tensor
            f = f.view(f.size(0), -1)
            #print(f.size())
            f = self.fc_pre(f)

            # stores the tensor f in the shape [fs.size(0), fs.size(2)] to each “row” in fs to prepare for learning the sequence in RNN
            fs[:, i, :] = f
            # print('fs')
            # print(fs.size())
       
            
        outputs, hidden = self.rnn(fs, hidden)
        #print(outputs.size())
        
        #the training code applies crossentropyloss as criterion which also applies softmax to the output so I don't have to use softmax here
        outputs = self.fc(outputs)
        #outputs = self.softmax(self.fc(outputs))
        #print(outputs)
        #print(outputs.size())
        
        return outputs

the fc_size is equivalent to the input to my MLP which is 201. I wish to know if I am initializing the RNN correctly. Thank you in advance