我很难将 albert(更具体地说，albert_base模型(转换为 tflite。这是我使用 bert-for-tf2 (https://github.com/kpe/bert-for-tf2(<定义我的模型的代码 _x002D_=">
import tensorflow as tf
import numpy as np
from tensorflow.keras.layers import Input, Flatten, AveragePooling1D
from tensorflow.keras.models import Model
import bert
import sentencepiece as spm
def load_pretrained_albert():
model_name = "albert_base"
albert_dir = bert.fetch_tfhub_albert_model(model_name, ".models")
model_params = bert.albert_params(model_name)
l_bert = bert.BertModelLayer.from_params(model_params, name="albert")
# use in Keras Model here, and call model.build()
max_seq_len = 128
l_input_ids = Input(shape=(max_seq_len,), dtype='float32', name="l_input_ids")
output = l_bert(l_input_ids)
pooled_output = AveragePooling1D(pool_size=max_seq_len, data_format="channels_last")(output)
pooled_output = Flatten()(pooled_output) # poooled_output: [batch_size, embedding_dimension=768]
model = Model(inputs=[l_input_ids], outputs=[pooled_output])
model.build(input_shape=(None, max_seq_len))
bert.load_albert_weights(l_bert, albert_dir)
return model
但是当我尝试使用以下代码将模型转换为 tflite 时，
converter = tf.lite.TFLiteConverter.from_keras_model(m)
tflite_model = converter.convert()
发生以下错误：
File "C:UsershygkiAnaconda3libsite-packagestensorflow_corelitepythonlite.py", line 405, in convert
self._funcs[0], lower_control_flow=False)
File "C:UsershygkiAnaconda3libsite-packagestensorflow_corepythonframeworkconvert_to_constants.py", line 575, in convert_variables_to_constants_v2
converted_input_indices)
File "C:UsershygkiAnaconda3libsite-packagestensorflow_corepythonframeworkconvert_to_constants.py", line 371, in _construct_concrete_function
new_output_names)
File "C:UsershygkiAnaconda3libsite-packagestensorflow_corepythoneagerwrap_function.py", line 620, in function_from_graph_def
wrapped_import = wrap_function(_imports_graph_def, [])
File "C:UsershygkiAnaconda3libsite-packagestensorflow_corepythoneagerwrap_function.py", line 598, in wrap_function
collections={}),
File "C:UsershygkiAnaconda3libsite-packagestensorflow_corepythonframeworkfunc_graph.py", line 915, in func_graph_from_py_func
func_outputs = python_func(*func_args, **func_kwargs)
File "C:UsershygkiAnaconda3libsite-packagestensorflow_corepythoneagerwrap_function.py", line 83, in __call__
return self.call_with_variable_creator_scope(self._fn)(*args, **kwargs)
File "C:UsershygkiAnaconda3libsite-packagestensorflow_corepythoneagerwrap_function.py", line 89, in wrapped
return fn(*args, **kwargs)
File "C:UsershygkiAnaconda3libsite-packagestensorflow_corepythoneagerwrap_function.py", line 618, in _imports_graph_def
importer.import_graph_def(graph_def, name="")
File "C:UsershygkiAnaconda3libsite-packagestensorflow_corepythonutildeprecation.py", line 507, in new_func
return func(*args, **kwargs)
File "C:UsershygkiAnaconda3libsite-packagestensorflow_corepythonframeworkimporter.py", line 405, in import_graph_def
producer_op_list=producer_op_list)
File "C:UsershygkiAnaconda3libsite-packagestensorflow_corepythonframeworkimporter.py", line 505, in _import_graph_def_internal
raise ValueError(str(e))
ValueError: Input 0 of node model/albert/embeddings/word_embeddings/embedding_lookup was passed float from model/albert/embeddings/word_embeddings/embedding_lookup/Read/ReadVariableOp/resource:0 incompatible with expected resource.
因此，我尝试以saved_model格式保存模型，并使用以下代码重试转换：
converter = tf.lite.TFLiteConverter.from_saved_model('saved_model_path')
tflite_model = converter.convert()
但是，再次出现相同的错误消息。
File "C:UsershygkiAnaconda3libsite-packagestensorflow_corelitepythonlite.py", line 405, in convert
self._funcs[0], lower_control_flow=False)
File "C:UsershygkiAnaconda3libsite-packagestensorflow_corepythonframeworkconvert_to_constants.py", line 575, in convert_variables_to_constants_v2
converted_input_indices)
File "C:UsershygkiAnaconda3libsite-packagestensorflow_corepythonframeworkconvert_to_constants.py", line 371, in _construct_concrete_function
new_output_names)
File "C:UsershygkiAnaconda3libsite-packagestensorflow_corepythoneagerwrap_function.py", line 620, in function_from_graph_def
wrapped_import = wrap_function(_imports_graph_def, [])
File "C:UsershygkiAnaconda3libsite-packagestensorflow_corepythoneagerwrap_function.py", line 598, in wrap_function
collections={}),
File "C:UsershygkiAnaconda3libsite-packagestensorflow_corepythonframeworkfunc_graph.py", line 915, in func_graph_from_py_func
func_outputs = python_func(*func_args, **func_kwargs)
File "C:UsershygkiAnaconda3libsite-packagestensorflow_corepythoneagerwrap_function.py", line 83, in __call__
return self.call_with_variable_creator_scope(self._fn)(*args, **kwargs)
File "C:UsershygkiAnaconda3libsite-packagestensorflow_corepythoneagerwrap_function.py", line 89, in wrapped
return fn(*args, **kwargs)
File "C:UsershygkiAnaconda3libsite-packagestensorflow_corepythoneagerwrap_function.py", line 618, in _imports_graph_def
importer.import_graph_def(graph_def, name="")
File "C:UsershygkiAnaconda3libsite-packagestensorflow_corepythonutildeprecation.py", line 507, in new_func
return func(*args, **kwargs)
File "C:UsershygkiAnaconda3libsite-packagestensorflow_corepythonframeworkimporter.py", line 405, in import_graph_def
producer_op_list=producer_op_list)
File "C:UsershygkiAnaconda3libsite-packagestensorflow_corepythonframeworkimporter.py", line 505, in _import_graph_def_internal
raise ValueError(str(e))
ValueError: Input 0 of node StatefulPartitionedCall/model/albert/embeddings/word_embeddings/embedding_lookup was passed float from Func/StatefulPartitionedCall/input/_2:0 incompatible with expected resource.
所以我的理解是，当预期的数据类型不是浮点数时，embedding_lookup会用浮点数馈送。但是预期的数据类型是什么？有没有办法让我知道？另外，有解决此问题的方法吗？
任何帮助我将albert_base转换为 tflite formtat
！关于"IdentityN"错误，您是否尝试用SELECT_TF_OPS进行转换？ https://www.tensorflow.org/lite/guide/ops_select
有趣的是，我已经为这个问题苦苦挣扎了几个小时，但只有在我上传问题后我才解决了问题......
所以解决方案是，使用张量流版本 1.15.0！ 使用 tensorflow2 似乎会导致此问题。
但是，我仍然无法将模型转换为 tflite，因为它尚不支持"IdentityN"操作。我不认为我可以自己编写自定义操作，所以我认为我应该等待 tflite 更新......
使用官方存储库中的 ALBERT 2.0 (tf 2.0( 模型。将 https://github.com/google-research/ALBERT/blob/master/modeling.py#L516 更改为tf.gather(tf.identity(embedding_table), input_ids)。然后尝试像以前一样使用 tflite 进行转换。如果没有，请在此处发表评论。