python - 遇到 Unresolved 自定义操作 : TensorListFromTensor. 节点号 4 (TensorListFromTensor) 准备失败

Question

我是 tensorflow 和机器学习的新手。在这里，我尝试创建自己的文本分类。我面临以下问题。我在 Android 上加载 tflite 模型时遇到以下错误。

 Process: org.tensorflow.lite.examples.textclassification, PID: 12031
java.lang.IllegalStateException: Internal error: Unexpected failure when preparing tensor allocations: Encountered unresolved custom op: TensorListFromTensor.
Node number 4 (TensorListFromTensor) failed to prepare.

    at org.tensorflow.lite.NativeInterpreterWrapper.allocateTensors(Native Method)
    at org.tensorflow.lite.NativeInterpreterWrapper.init(NativeInterpreterWrapper.java:96)
    at org.tensorflow.lite.NativeInterpreterWrapper.<init>(NativeInterpreterWrapper.java:61)
    at org.tensorflow.lite.Interpreter.<init>(Interpreter.java:224)
    at org.tensorflow.lite.Interpreter.<init>(Interpreter.java:182)
    at org.tensorflow.lite.examples.textclassification.TextClassificationClient.loadModel(TextClassificationClient.java:139)
    at org.tensorflow.lite.examples.textclassification.TextClassificationClient.load(TextClassificationClient.java:129)
    at org.tensorflow.lite.examples.textclassification.MainActivity.lambda$onStart$1$MainActivity(MainActivity.java:65)
    at org.tensorflow.lite.examples.textclassification.-$$Lambda$MainActivity$3O7XdyFVukGQ46LZyOIpHdTRepg.run(Unknown Source:2)
    at android.os.Handler.handleCallback(Handler.java:873)
    at android.os.Handler.dispatchMessage(Handler.java:99)
    at android.os.Looper.loop(Looper.java:193)
    at android.app.ActivityThread.main(ActivityThread.java:6898)
    at java.lang.reflect.Method.invoke(Native Method)
    at com.android.internal.os.RuntimeInit$MethodAndArgsCaller.run(RuntimeInit.java:537)
    at com.android.internal.os.ZygoteInit.main(ZygoteInit.java:858)

Py 笔记本代码

input = tf.keras.layers.Input(shape=(max_len,))
x = tf.keras.layers.Embedding(max_words, embed_size, weights=[embedding_matrix], trainable=False)(input)

x = tf.keras.layers.Bidirectional(tf.keras.layers.GRU(128, return_sequences=True, dropout=0.1,
                                                      recurrent_dropout=0.1))(x)

x = tf.keras.layers.Conv1D(64, kernel_size=3, padding="valid", kernel_initializer="glorot_uniform")(x)

avg_pool = tf.keras.layers.GlobalAveragePooling1D()(x)
max_pool = tf.keras.layers.GlobalMaxPooling1D()(x)

x = tf.keras.layers.concatenate([avg_pool, max_pool])

preds = tf.keras.layers.Dense(6, activation="sigmoid")(x)

model = tf.keras.Model(input, preds)

model.summary()

在 python 中运行下面的代码也会导致相同的错误

interpreter = tf.lite.Interpreter(model_path="xxx.tflite")
interpreter.allocate_tensors()

Error below
---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
<ipython-input-20-988ba8e1e2e5> in <module>
      1 interpreter = tf.lite.Interpreter(model_path="touch_tf.tflite")
----> 2 interpreter.allocate_tensors()

/usr/local/lib/python3.7/site-packages/tensorflow_core/lite/python/interpreter.py in allocate_tensors(self)
    240   def allocate_tensors(self):
    241     self._ensure_safe()
--> 242     return self._interpreter.AllocateTensors()
    243 
    244   def _safe_to_run(self):

/usr/local/lib/python3.7/site-packages/tensorflow_core/lite/python/interpreter_wrapper/tensorflow_wrap_interpreter_wrapper.py in AllocateTensors(self)
    108 
    109     def AllocateTensors(self):
--> 110         return _tensorflow_wrap_interpreter_wrapper.InterpreterWrapper_AllocateTensors(self)
    111 
    112     def Invoke(self):

RuntimeError: Encountered unresolved custom op: TensorListFromTensor.Node number 4 (TensorListFromTensor) failed to prepare.

Answer 1

此链接帮助我解决了此问题。

Git Issue #29472

设置converter.experimental_new_converter =True会有帮助