我正在尝试使用具有embeddings_initializer = keras.initializers.Identity的keras.layers.Embedding层构建Keras模型。
这应该让我直接在Keras模型中实现单热编码，同时为模型提供单词列表的嵌入表示（即从单词键到稀疏二进制向量）
每个样本的输入示例：
[ [5,22,3], # Statement encoding
[6,9,1,76], # Context encoding
]
我正在尝试实现Keras One Hot Encoding Memory Management - best Possible way out
中描述的内容（看但以理的答案）
嵌入层的预期输出：
[ [ [B11], [B12], [B13] ], # Statement encoding for lstm_emb_phrase
[[B21], [B22], [B23], [B24] ], # Context encoding for lstm_emb_cont
]
其中＃Bij＃应该是短语i的单词j的二进制数组。
这是我当前编写的代码：
DEFAULT_INNER_ACTIVATION = 'relu'
DEFAULT_OUTPUT_ACTIVATION = 'softplus'
def __init__(self, sentence_max_lenght, ctx_max_len, dense_features_dim, vocab_size):
lstm_input_phrase = keras.layers.Input(shape=(sentence_max_lenght,), name='L0_STC_MyApp')
lstm_input_cont = keras.layers.Input(shape=(ctx_max_len,), name='L0_CTX_MyApp')
# The following line is #56
lstm_emb_phrase = keras.layers.Embedding(vocab_size, vocab_size, embeddings_initializer=keras.initializers.Identity,
input_length=sentence_max_lenght, name='L0E_STC_MyApp')(lstm_input_phrase)
lstm_emb_phrase = keras.layers.LSTM(DEFAULT_MODEL_L1_STC_DIM, name='L1_STC_MyApp')(lstm_emb_phrase)
lstm_emb_phrase = keras.layers.Dense(DEFAULT_MODEL_L2_STC_DIM, name='L2_STC_MyApp', activation=DEFAULT_INNER_ACTIVATION)(lstm_emb_phrase)
lstm_emb_cont = keras.layers.Embedding(vocab_size, vocab_size,
embeddings_initializer=keras.initializers.Identity,
input_length=ctx_max_len, name='L0E_CTX_MyApp')(lstm_input_cont)
lstm_emb_cont = keras.layers.LSTM(DEFAULT_MODEL_L1_CTX_DIM, name='L1_CTX_MyApp')(lstm_emb_cont)
lstm_emb_cont = keras.layers.Dense(DEFAULT_MODEL_L2_CTX_DIM, name='L2_CTX_MyApp', activation=DEFAULT_INNER_ACTIVATION)(lstm_emb_cont)
x = keras.layers.concatenate([lstm_emb_phrase, lstm_emb_cont])
x = keras.layers.Dense(DEFAULT_MODEL_L3_DIM, activation=DEFAULT_INNER_ACTIVATION)(x)
x = keras.layers.Dense(DEFAULT_MODEL_L4_DIM, activation=DEFAULT_INNER_ACTIVATION)(x)
main_output = keras.layers.Dense(DEFAULT_OUTPUT_DIM, activation=DEFAULT_OUTPUT_ACTIVATION)(x)
self.model = keras.models.Model(inputs=[lstm_input_phrase, lstm_input_cont],
outputs=main_output)
self.model.compile(loss='binary_crossentropy', metrics=['accuracy'])
这是我尝试运行脚本时遇到的错误：
WARNING:tensorflow:From [PATH_TO_MyApp]\venv\lib\site-packages\tensorflow\python\framework\op_def_library.py:263: colocate_with (
from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.
Instructions for updating:
Colocations handled automatically by placer.
Traceback (most recent call last):
File "my_script.py", line 307, in <module>
main()
File "my_script.py", line 60, in main
flag = flag or train_r_skills(argsparser)
File "my_script.py", line 230, in train_r_skills
sk_extr = MyKerasModel(stm_max_len, ctx_max_len, segm_max_len, vocab_size)
File "[PATH_TO_MyApp]\MyApp\parser\MyKerasModel_002.py", line 56, in __init__
input_length=ctx_features_dim, name='L0E_CTX_MyApp')(lstm_input_cont)
File "[PATH_TO_MyApp]\venv\lib\site-packages\keras\engine\base_layer.py", line 431, in __call__
self.build(unpack_singleton(input_shapes))
File "[PATH_TO_MyApp]\venv\lib\site-packages\keras\layers\embeddings.py", line 109, in build
dtype=self.dtype)
File "[PATH_TO_MyApp]\venv\lib\site-packages\keras\legacy\interfaces.py", line 91, in wrapper
return func(*args, **kwargs)
File "[PATH_TO_MyApp]\venv\lib\site-packages\keras\engine\base_layer.py", line 252, in add_weight
constraint=constraint)
File "[PATH_TO_MyApp]\venv\lib\site-packages\keras\backend\tensorflow_backend.py", line 402, in variable
v = tf.Variable(value, dtype=tf.as_dtype(dtype), name=name)
File "[PATH_TO_MyApp]\venv\lib\site-packages\tensorflow\python\ops\variables.py", line 213, in __call__
return cls._variable_v1_call(*args, **kwargs)
File "[PATH_TO_MyApp]\venv\lib\site-packages\tensorflow\python\ops\variables.py", line 176, in _variable_v1_call
aggregation=aggregation)
File "[PATH_TO_MyApp]\venv\lib\site-packages\tensorflow\python\ops\variables.py", line 155, in <lambda>
previous_getter = lambda **kwargs: default_variable_creator(None, **kwargs)
File "[PATH_TO_MyApp]\venv\lib\site-packages\tensorflow\python\ops\variable_scope.py", line 2495, in default_variable_creator
expected_shape=expected_shape, import_scope=import_scope)
File "[PATH_TO_MyApp]\venv\lib\site-packages\tensorflow\python\ops\variables.py", line 217, in __call__
return super(VariableMetaclass, cls).__call__(*args, **kwargs)
File "[PATH_TO_MyApp]\venv\lib\site-packages\tensorflow\python\ops\variables.py", line 1395, in __init__
constraint=constraint)
File "[PATH_TO_MyApp]\venv\lib\site-packages\tensorflow\python\ops\variables.py", line 1503, in _init_from_args
initial_value(), name="initial_value", dtype=dtype)
TypeError: __call__() missing 1 required positional argument: 'shape'
您能帮我了解发生了什么吗？
编辑： 我也按照C.Lightfoot的建议检查了Convert code from Keras 1 to Keras 2: TypeError: __call__() missing 1 required positional argument: 'shape'，但是，据我所知，这似乎与我的问题不符（我没有将代码从Keras 1转换为Keras 2）。如果我缺少有关本文的任何内容，请给我更多提示。
干杯， / H