Repository navigation
by usingin #1546
Description
Describe the bug
i am trying to deploy a pretrained keras model in amazon sage maker,
To reproduce
- I converted the keras model into tensorflow model format as sagemkaer supported one.
[Link: https://aws.amazon.com/blogs/machine-learning/deploy-trained-keras-or-tensorflow-models-using-amazon-sagemaker/] - I created a code folder and added inference.py and requirements.txt file
- The inference script as mentioned in this link, i created with some changes for my model.
Link : https://github.com/aws/sagemaker-tensorflow-serving-container#prepost-processing
`# inference.py
import tensorflow as tf
import numpy as np
import requests
import base64
from PIL import Image
import io
from collections import namedtuple
import cv2
import json
import os
JPEG_CONTENT_TYPE = 'application/x-image'
JSON_CONTENT_TYPE = 'application/json'
Context = namedtuple
def input_handler(data, context):
#if context.request_content_type == 'application/x-image':
if context== 'application/x-image':
#payload = data.read()
#encoded_image = base64.b64encode(payload).decode('utf-8')
image = Image.open(io.BytesIO(data))
result_image =cv2.cvtColor(np.array(image), cv2.COLOR_BGR2RGB)
img_cpy=cv2.resize(result_image, (300,300))
img_cpy=img_cpy/255.
img_cpy=np.expand_dims(img_cpy,0)
print(type(img_cpy))
encoded_image = base64.b64encode(img_cpy).decode('utf-8')
print(type(encoded_image))
instance = [{"b64": encoded_image}]
return json.dumps({"instances": instance})
else:
_return_error(415, 'Unsupported content type "{}"'.format(
context.request_content_type or 'Unknown'))
def output_handler(data, context):
response = json.loads(data.content.decode('utf-8'))
pred = response.get('predictions')
prediction=np.argmax(pred)
prediction = str(prediction)
return json.dumps({'classification_id':prediction})
def _return_error(code, message):
raise ValueError('Error: {}, {}'.format(str(code), message))
- i zipped (tared) the model folder into tar.gz format and uploaded in s3 bucket.
- Then i open a sagemaker notebooke instance to deply, i uploaded the code folder in the notebook instance also then i started to deploy it.
It show error :
2020-06-03T15:21:54.839+05:30 | /usr/local/lib/python3.5/dist-packages/tensorflow/python/framework/dtypes.py:518: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'. _np_qint16 = np.dtype([("qint16", np.int16, 1)])
2020-06-03T15:21:54.840+05:30 | Traceback (most recent call last): File "/usr/local/lib/python3.5/dist-packages/gunicorn/arbiter.py", line 583, in spawn_worker worker.init_process() File "/usr/local/lib/python3.5/dist-packages/gunicorn/workers/ggevent.py", line 203, in init_process super(GeventWorker, self).init_process() File "/usr/local/lib/python3.5/dist-packages/gunicorn/workers/base.py", line 129, in init_process self.load_wsgi() File "/usr/local/lib/python3.5/dist-packages/gunicorn/workers/base.py", line 138, in load_wsgi self.wsgi = self.app.wsgi() File "/usr/local/lib/python3.5/dist-packages/gunicorn/app/base.py", line 67, in wsgi self.callable = self.load() File "/usr/local/lib/python3.5/dist-packages/gunicorn/app/wsgiapp.py", line 52, in load return self.load_wsgiapp() File "/usr/local/lib/python3.5/dist-packages/gunicorn/app/wsgiapp.py", line 41, in load_wsgiapp return util.import_app(self.app_uri) File "/usr/local/lib/python3.5/dist-packages/gunicorn/util.py", line 350, in import_app import(module) File "/sagemaker/python_service.py", line 239, in resources.add_routes(app) File "/sagemaker/python_service.py", line 227, in add_routes invocation_resource = InvocationResource() File "/sagemaker/python_service.py", line 48, in init self._handler, self._input_handler, self._output_handler = self._import_handlers() File "/sagemaker/python_service.py", line 68, in _import_handlers spec.loader.exec_module(inference) File "/opt/ml/model/code/inference.py", line 8, in import cv2 File "/usr/local/lib/python3.5/dist-packages/cv2/init.py", line 5, in from .cv2 import *
*Expected behavior
i need to check the inference, when i pass the image by using invoke endpoint api. it should return the json response.
System information
A description of your system. Please provide:
- SageMaker Python SDK version:
- Framework name (eg. PyTorch) or algorithm (eg. KMeans): tensorflow backed keras classification
- Framework version:1.14 or 1.12
- Python version:3
- CPU or GPU:GPU
- Custom Docker image (Y/N):N
Activity
nadiaya commentedon Jun 9, 2020
OpenCV is not installed by default in the container 1.14 gpu container.
Are you install it using requirements.txt? Can you share that file?
Also it looks like the actual error message is cut off. Can you share the full stack with the error message?
saisubramani commentedon Jun 9, 2020
hi,Thanks for the reply @nadiaya
I had installed opencv-python in requirements.txt file.
Below mentioned is the packages which we installed using the requirements.txt file.
Full Cloud watch error:
laurenyu commentedon Jul 23, 2020
sorry for the delayed response here.
looks like this is the error:
according to https://askubuntu.com/questions/1060903/importerror-libgthread-2-0-so-0-cannot-open-shared-object-file-no-such-file-o, there needs to be a Linux library installed in order to use OpenCV.
Unfortunately, there's not a way to do this at runtime, so what you'll have to do is use a Deep Learning Container as a base and install the necessary libraries.
DLC images: https://aws.amazon.com/releasenotes/available-deep-learning-containers-images