Hello everybody,
Recently, I am trying to implement a two-layer LSTM model and train it with my dataset.
This is my model implementation:
class LstmModel(nn.Module):
def __init__(self, device):
super(LstmModel, self).__init__()
self.lstm1 = nn.LSTM(
input_size=2048, hidden_size=1024, batch_first=True)
self.lstm2 = nn.LSTM(
input_size=1024, hidden_size=128, batch_first=True)
self.linear = nn.Linear(128, 1)
self.sigmoid = nn.Sigmoid()
self.device = device
def forward(self, x):
h0 = torch.zeros((1, 32, 1024)).to(self.device)
c0 = torch.zeros((1, 32, 1024)).to(self.device)
x, _ = self.lstm1(x, (h0, c0))
h1 = torch.zeros((1, 32, 128)).to(self.device)
c1 = torch.zeros((1, 32, 128)).to(self.device)
x, _ = self.lstm2(x, (h1, c1))
x = self.linear(x)
x = self.sigmoid(x)
return x
And this is my dataset implementation.
class FeaturesDataset(Dataset):
def __init__(self):
csvFile = pd.read_csv(
"E:\\Datasets\\VQA\\KoNVID_1k_LSTM_CNN\\Features_240\\raw_data.csv")
self.X = csvFile['flickr_id']
self.y = csvFile['mos']
def __len__(self):
return len(self.X)
def __getitem__(self, idx):
dist_path = 'E:\Datasets\VQA\KoNVID_1k_LSTM_CNN\Features_240'
video_name = str(int(self.X[idx])) + '.txt'
x = np.loadtxt(join(dist_path, video_name))
y = self.y[idx]
return torch.tensor(x), y
The following code is one row of my dataset.
The first element is my features, with the shape of 240 * 2048, which 240 is the length of my sequence and 2048 is the number of input features.
(tensor([[0.1705, 0.0173, 1.1980, ..., 0.0022, 1.1543, 0.0000],
[0.1560, 0.0206, 1.2292, ..., 0.0000, 1.1819, 0.0000],
[0.0830, 0.0018, 1.3242, ..., 0.0026, 1.3389, 0.0024],
...,
[0.0645, 0.0062, 1.3789, ..., 0.0031, 1.0773, 0.0000],
[0.0683, 0.0048, 1.3562, ..., 0.0025, 1.1218, 0.0000],
[0.0811, 0.0058, 1.4024, ..., 0.0017, 1.1745, 0.0000]],
dtype=torch.float64), 4.64)
(venv) PS C:\Users\HP\my-python-pro
The problem is while training the model, I get RuntimeError: Input and hidden tensors are not the same dtype, found input tensor with Double and hidden tensor with Float.
Also, the batch size of my data is 32. Does anybody know the reason?