我開發(fā)了一個遷移學(xué)習(xí)應(yīng)用程序，我正在為我的數(shù)據(jù)流重新訓(xùn)練 MobileNetV2。我正在使用來自 tensorflow-hub 的retrain.py重新訓(xùn)練模型并且沒有進(jìn)行任何修改。當(dāng)我從終端運行腳本時，在將模型下載到我的用戶配置文件中的臨時目錄后，我會直接收到此警告。Importing a graph with a lower producer version 26 into an existing graph with producer version 27. Shape inference will have run different parts of the graph with different producer versions.在調(diào)試過程中，我創(chuàng)建了一個test.py腳本來找出警告的來源：import tensorflow as tfimport tensorflow_hub as hubdef create_module_graph(module_spec):  """Creates a graph and loads Hub Module into it.  Args:    module_spec: the hub.ModuleSpec for the image module being used.  Returns:    graph: the tf.Graph that was created.    bottleneck_tensor: the bottleneck values output by the module.    resized_input_tensor: the input images, resized as expected by the module.    wants_quantization: a boolean, whether the module has been instrumented      with fake quantization ops.  """  FAKE_QUANT_OPS = ('FakeQuantWithMinMaxVars',                     'FakeQuantWithMinMaxVarsPerChannel')  height, width = hub.get_expected_image_size(module_spec)  with tf.Graph().as_default() as graph:    resized_input_tensor = tf.placeholder(tf.float32, [None, height, width, 3])    m = hub.Module(module_spec)    bottleneck_tensor = m(resized_input_tensor)    wants_quantization = any(node.op in FAKE_QUANT_OPS                             for node in graph.as_graph_def().node)  return graph, bottleneck_tensor, resized_input_tensor, wants_quantizationdef main():   module_spec = hub.load_module_spec('https://tfhub.dev/google/imagenet/mobilenet_v2_100_96/classification/2')   graph, bottleneck_tensor, resized_input_tensor, wants_quantization = create_module_graph(module_spec)if __name__ =='__main__':   main()并發(fā)現(xiàn)它起源create_module_graph于retrain.py. 當(dāng)我使用從終端運行腳本時python test.py，我從上面收到生產(chǎn)者警告。但是，當(dāng)我main()從 ipython 控制臺運行時，我沒有收到生產(chǎn)者版本警告。當(dāng)我所做的只是從 tensorflow-hub 存儲庫創(chuàng)建圖表時，我不確定為什么會發(fā)生這種情況。我查看了版本兼容性文檔，沒有看到與錯誤特別相關(guān)的任何內(nèi)容。查看源代碼，似乎表明我的圖形在構(gòu)建之前已減少到最低版本。這是需要擔(dān)心的嗎？它會改變您加載圖表以進(jìn)行預(yù)測的方式嗎？