Stack Overflow
Stack Overflow

Questions tagged [mlops]

This tag is for programming questions about MLOps, which is the application of DevOps principles in the design and deployment of Machine Learning (ML) systems.

See also:

Related tags

  • mlflow
  • kubeflow
  • feature-store
228 questions
0
votes
0 answers

AWS Sagemaker inference directly via DB (without using S3 for data)

I'm learning to use AWS SageMaker for MLops. I have a trained model (saved via joblib as a .sav file) on S3. I also published it in the model registry. I have a simple Python script for inference: it imports the model into the virtual machine, then…
Leo
  • 772
  • 1
  • 6
  • 15
0
votes
0 answers

Is there a need to create a cluster on Databricks if I need to use the workspace as just the model registry using MLFlow

I need to use the Databricks workspace as a model registry where I would maintain the lifecycle of my model, and in the other 3 Ddatabricks workspaces I will manage as Dev, Stage and Production for my MLOps pipeline. Do I need to create a compute…
Shivam Sharma
  • 17
  • 4
0
votes
1 answer

Issue with MLRun Spark service start and impact all Jupyter notebooks

I reconfigured Spark infrastructure in K8s (as part of MLRun/Iguazio platform) and after that, I got a lot of issues in level of services: Spark service (with information Failed) All jupyter notebooks (with information Failed dependencies) and…
JIST
  • 1,139
  • 2
  • 8
  • 30
0
votes
0 answers

Running the code gives "upstream request timeout", where me Iris data set has around 300,000 lines of data and I my rest api is from Azure cloud

I have even tried reducing the length of data, but it's not working import requests import time import json with open("Iris.csv") as f: line = f.readline() headers = line.strip().split(",") features = [] for l in…
deepthi
  • 1
  • 1
0
votes
1 answer

Issue with ingest values, 2x more

When I ingested values to the feature set, the pipeline was called 2x more (I used MLRun version 1.2.1). It seems as the issue, do you know why? I used this code: import mlrun import mlrun.feature_store as fstore # mlrun: start-code import…
JIST
  • 1,139
  • 2
  • 8
  • 30
0
votes
0 answers

AttributeError: 'NoneType' object has no attribute 'encode' and AttributeError: 'InferenceServerClient' object has no attribute '_stream'

I had two 2 docker container in the server. One is Triton Client Server whose GRPC port I set is 1747. Triton Client Server port had a TorchScript model running on it. The other container is where I want to call grpcclient.InferenceServerClient to…
Văn Tuấn Nguyễn
  • 11
  • 2
0
votes
1 answer

How to load private python package when loading a MLFlow model?

I am trying to use a private Python package as a model using the mlflow.pyfunc.PythonModel. My conda.yaml looks like channels: - defaults dependencies: - python=3.10.4 - pip - pip: - mlflow==2.1.1 - pandas - --extra-index-url…
Vikrant Yadav
  • 303
  • 3
  • 18
0
votes
1 answer

Scaling endpoint that uses ML models in AWS

I have an API (python+FastAPI) with different endpoints. One of these endpoints manages images, and apply different models (up to 4) one after another one. Furthermore, there can be a preprocessing and postprocessing between these models are applied…
Learning from masters
  • 2,032
  • 3
  • 29
  • 42
0
votes
0 answers

Try deploying a custom ML model with an endpoint by using a custom image on Google Cloud's Vertex AI

I have been banging my head around this for a while and Google Cloud does not have a lot of documentation about this issue. What I am trying to do is deploy a custom ML model on Google Cloud Vertex by: Uploading the model onto the Model Registry in…
Neil Bhutada
  • 1
  • 2
0
votes
1 answer

Exception during ingestion, MLRunAccessDeniedError

I got this issue during data ingest/write: ~/.pythonlibs/jist-jupyter/lib/python3.7/site-packages/mlrun/errors.py in raise_for_status(response, message) 82 raise STATUS_ERRORS[response.status_code]( 83 …
JIST
  • 1,139
  • 2
  • 8
  • 30
0
votes
0 answers

mlflow problems with metrics logging

everyone. I'm starting to work with mlflow, and right now i'm tryinng to implement it into a project but it isn't working properly. The mlflow.log_metric()/mlflow.log_metrics module isn't working: # creating the context with…
Lucas NG
  • 1
  • 2
0
votes
1 answer

Redshift removal of explicit locks lead to missing rows/duplicated rows. Best resolution?

I have a use case where some of our redshift tables are used by multiple data scientists at the same time for tuning. If they are tuning at the same time and import data at the same time, we end up with missing rows and duplicated rows. Awhile ago,…
jimhens
  • 1
  • 1
0
votes
1 answer

AWS MLOps - Do we need dev/test environments for the model-development environment that data scientist work on (SageMaker))

assuming I have AWS accounts: DEV (where data scientists use SageMaker notebooks/studio to actively explore data and develop models) Test (where the model monitor happens) Prod (where the accepted model is hosted) My question is, from engineering…
wawawa
  • 2,835
  • 6
  • 44
  • 105
0
votes
1 answer

Storing encoded Features in the Feature Store?

I am currently in the design phase of a feature store pipeline to train and serve ML models. One thing which confuses me is the "state" in which features are supposed to be stored in the feature store. Let's assume I have a data source which has an…
dosmarder
  • 1
0
votes
0 answers

Kedro register dataset from a board from pins package

For my project I want to use a combination of kedro for the pipeline orchestration and pins for data and model versioning. I have some data which I stored on a board from the pins package. As I have multiple versions, I am not sure how to specify…
Mischa
  • 137
  • 8
1 2 3
…
15 16