loss stuck at 0.69 in word2vec lstm binary classification

Question

i have a data with 4800 size balanced 2 class. when i training it the results show loss : 0.6932 and accuracy around 49-50% per epoch. here is my vector word2vec convert code :

vector_size = 100
w2v_weight_matrix = np.zeros((vocab_size ,vector_size))
w2v_weight_matrix.shape

for word, index in tokenizer.word_index.items():
    if index < vocab_size: # since index starts with zero 
        if word in id_w2v.wv.key_to_index:
            w2v_weight_matrix[index] = id_w2v.wv[word]
        else:
            w2v_weight_matrix[index] = np.zeros(100)

and here is my word2vec-lstm structure code :

model=Sequential()
    model.add(Embedding(vocab_size, embed_dim, input_length=max_len, embeddings_initializer = Constant(w2v_weight_matrix), trainable = False))
    model.add(Dropout(0.2))
    model.add(LSTM(units=100))
    model.add(Dropout(0.2))
    model.add(Dense(1,activation='sigmoid'))
    model.compile(loss='binary_crossentropy',
                  optimizer='adam',
                  metrics=['accuracy'])

is anyone know what my mistakes and know the solution here? :( pardon my english..