我使用tensorflow服务部署自己的模型。通常它可以正常工作,但有时会收到这样的错误消息:
AbortionError(code=StatusCode.UNAVAILABLE, details="OS Error")
我不知道为什么,有人可以帮助我吗?
我的代码:
channel = implementations.insecure_channel(tfserving_food_cls_host, int(tfserving_foodcls_port))
stub = prediction_service_pb2.beta_create_PredictionService_stub(channel)
serving_request = predict_pb2.PredictRequest()
serving_request.model_spec.name = 'default'
serving_request.model_spec.signature_name = 'predict_images'
serving_request.inputs['images'].CopyFrom(
tf.contrib.util.make_tensor_proto(image299,
shape=[1, 299, 299, 3]))
serving_results = stub.Predict(serving_request, 10)
答案 0 :(得分:0)
这看起来像操作系统错误,如日志所述?您尝试过其他主机吗?