'TFEmbeddings' object has no attribute 'word_embeddings'

[hossein-amirkhani]

Trying to run this example, I encountered 'TFEmbeddings' object has no attribute 'word_embeddings' error. Any help is appreciated.

[dr-smgad]

Try to use '.weight' instead of .word_embeddings as per hugging face latest implementation:

`class TFBertEmbeddings(tf.keras.layers.Layer):
"""Construct the embeddings from word, position and token_type embeddings."""

def __init__(self, config: BertConfig, **kwargs):
    super().__init__(**kwargs)

    self.vocab_size = config.vocab_size
    self.type_vocab_size = config.type_vocab_size
    self.hidden_size = config.hidden_size
    self.max_position_embeddings = config.max_position_embeddings
    self.initializer_range = config.initializer_range
    self.embeddings_sum = tf.keras.layers.Add()
    self.LayerNorm = tf.keras.layers.LayerNormalization(epsilon=config.layer_norm_eps, name="LayerNorm")
    self.dropout = tf.keras.layers.Dropout(rate=config.hidden_dropout_prob)

def build(self, input_shape: tf.TensorShape):
    **with tf.name_scope("word_embeddings"):
        self.weight = self.add_weight**(
            name="weight",
            shape=[self.vocab_size, self.hidden_size],
            initializer=get_initializer(self.initializer_range),
        )`

[hossein-amirkhani]

Thanks. Any point on the new raised error: "TypeError: Cannot convert 'logits' to EagerTensor of dtype float"?