加載模型后我無法訪問圖層。我創建的模型如下：def create_model(vocab_dim, hidden_dim):    input_seq_axis1 = Axis('inputAxis1')    input_sequence_before = sequence.input_variable(shape=vocab_dim, sequence_axis=input_seq_axis1, is_sparse = use_sparse)    input_sequence_after = sequence.input_variable(shape=vocab_dim, sequence_axis=input_seq_axis1, is_sparse = use_sparse)    e=Sequential([        C.layers.Embedding(hidden_dim),        Stabilizer()        ],name='Embedding')    a = Sequential([        e,          C.layers.Recurrence(C.layers.LSTM(hidden_dim//2),name='ForwardRecurrence'),        ],name='ForwardLayer')    b = Sequential([        e,          C.layers.Recurrence(C.layers.LSTM(hidden_dim//2),go_backwards=True),       ],name='BackwardLayer')    latent_vector = C.splice(a(input_sequence_before), b(input_sequence_after))    bias = C.layers.Parameter(shape = (vocab_dim, 1), init = 0, name='Bias')    weights = C.layers.Parameter(shape = (vocab_dim, hidden_dim), init = C.initializer.glorot_uniform(), name='Weights')    z = C.times_transpose(weights, latent_vector,name='Transpose') + bias    z = C.reshape(z, shape = (vocab_dim))    return z然后我加載模型：def load_my_model(vocab_dim, hidden_dim):    z=load_model("models/lm_epoch0.dnn")    input_sequence_before = z.arguments[0]    input_sequence_after = z.arguments[1]    a=z.ForwardLayer    b=z.BackwardLayer    latent_vector = C.splice(a(input_sequence_before), b(input_sequence_after))我收到一個錯誤：TypeError("argument ForwardRecurrence 的類型 SequenceOver[inputAxis1][Tensor[100]] 與傳遞的變量的類型 SequenceOver[inputAxis1][SparseTensor[50000]] 不兼容",)看起來名稱引用的層 (z.ForwardLayer) 表示來自層立即輸入的函數。如何計算“latent_vector”（我需要這個變量來創建交叉熵和損失函數以繼續訓練）？