燒瓶模型部署的 Sagemaker 超時

Question

下面是 ECR Container 中的 predict.py。 Sagemaker 端點在重試 10-12 分鍾后給出“狀態：失敗”輸出。 /ping 和 /invocations 方法都可用

/opt/ml/code/predict.py
----------
logger = logging.getLogger()
logger.setLevel(logging.INFO)
classpath =  <.pkl file> 
model = pickle.load(open(classpath, "rb"))


app = flask.Flask(__name__)
print(app)

@app.route("/ping", methods=["GET"]
def ping():
    """Determine if the container is working and healthy."""
    return flask.Response(response="Flask running", status=200, mimetype="application/json")

@app.route("/invocations", methods=["POST"])
    ""InferenceCode""
    return flask.Response(response="Invocation Completed", status=200, 
    mimetype="application/json")

Below snippet was both added and removed , however I still have the endpoint in failed status

 if __name__ == '__main__':
     app.run(host='0.0.0.0',port=5000)

Error : 
"The primary container for production variant <modelname> did not pass the ping health check. Please check CloudWatch logs for this endpoint."


Sagemaker endpoint Cloudwatch logs.
[INFO] Starting gunicorn 20.1.0
[INFO] Listening at: http://0.0.0.0:8000 (1)
[INFO] Using worker: sync
[INFO] Booting worker with pid: 11```

Answer 1

您的預測器文件旨在測試模型是否已加載到 /ping 中，以及您是否可以在 /invocations 中執行推理。 如果您已在 SageMaker 上訓練您的模型，您需要按如下方式從 /opt/ml 目錄加載它。

prefix = "/opt/ml/"
model_path = os.path.join(prefix, "model")

class ScoringService(object):
    model = None  # Where we keep the model when it's loaded

    @classmethod
    def get_model(rgrs):
        """Get the model object for this instance, loading it if it's not already loaded."""
        if rgrs.model == None:
            with open(os.path.join(model_path, "rf-model.pkl"), "rb") as inp:
                rgrs.model = pickle.load(inp)
        return rgrs.model

    @classmethod
    def predict(rgrs, input):
        """For the input, do the predictions and return them.
        Args:
            input (a pandas dataframe): The data on which to do the predictions. There will be
                one prediction per row in the dataframe"""
        rf = rgrs.get_model()
        return rf.predict(input)

該類有助於加載您的模型，然后我們可以在 /ping 中進行驗證。

# The flask app for serving predictions
app = flask.Flask(__name__)


@app.route("/ping", methods=["GET"])
def ping():
    """Determine if the container is working and healthy. In this sample container, we declare
    it healthy if we can load the model successfully."""
    health = ScoringService.get_model() is not None  # You can insert a health check here

    status = 200 if health else 404
    return flask.Response(response="\n", status=status, mimetype="application/json")

在這里 SageMaker 將測試您是否正確加載了模型。 對於 /invocations 包括您傳遞到模型預測功能的任何數據格式的推理邏輯。

@app.route("/invocations", methods=["POST"])
def transformation():
    
    data = None

    # Convert from CSV to pandas
    if flask.request.content_type == "text/csv":
        data = flask.request.data.decode("utf-8")
        s = io.StringIO(data)
        data = pd.read_csv(s, header=None)
    else:
        return flask.Response(
            response="This predictor only supports CSV data", status=415, mimetype="text/plain"
        )

    print("Invoked with {} records".format(data.shape[0]))

    # Do the prediction
    predictions = ScoringService.predict(data)

    # Convert from numpy back to CSV
    out = io.StringIO()
    pd.DataFrame({"results": predictions}).to_csv(out, header=False, index=False)
    result = out.getvalue()
    
    
    return flask.Response(response=result, status=200, mimetype="text/csv")

確保如上所示設置或配置您的 predictor.py，以便 SageMaker 能夠正確理解/檢索您的模型。

我為 AWS 工作，我的意見是我自己的。