- 4626 Views
- 3 replies
- 3 kudos
Load a pyfunc model logged with Feature Store
Hi, I'm using Databricks Feature Store to register a custom model using a model wrapper as follows: # Log custom model to MLflow fs.log_model( artifact_path="model", model = production_model, flavor = mlflow.pyfunc, training_set = training_s...
- 4626 Views
- 3 replies
- 3 kudos
- 3 kudos
Hi @SOlivero Make sure that the model was in fact saved with the provided URI.The latest keyword will retrieve the latest version of the registered model when mlflow.pyfunc.load_model('models:/model_name/latest') is executed, not the highest version....
- 3 kudos
- 2112 Views
- 2 replies
- 3 kudos
Resolved! Hyperopt Ray integration
Hello,Is there a way to integrate Hyperopt with Ray parallelisation? I have a simulation framework which I want to optimise, and each simulation run is set up to be a Ray process, however I am calling one simulation run in the objective function. Thi...
- 2112 Views
- 2 replies
- 3 kudos
- 3 kudos
Hi @EmirHodzic Thank you for posting your question in the Databricks community. You can use Ray Tune, a tuning library that integrates with Ray, to parallelize your Hyperopt trials across multiple nodes.Here's a link to the documentation for HyperOpt...
- 3 kudos
- 4407 Views
- 1 replies
- 0 kudos
Provisioned concurrency of serving endpoints scales to zero
Hi, We provisioned the endpoint with 4 DBUs and also disabled the scale_to_zero option. For some reason, it randomly drops to 0 provisioned concurrency. Logs available in the serving endpoint service are not insightful. Currently, we are provisioning...
- 4407 Views
- 1 replies
- 0 kudos
- 0 kudos
Hi,I apologize if my question wasn't clear; let me clarify it.We are not using the scale_to_zero option and we are not doing any warmup requests so it should never scale to zero despite traffic or zero traffic right?
- 0 kudos
- 2882 Views
- 4 replies
- 1 kudos
Unable to access python variables in-between shells in same notebook
Unable to access Python variables in-between shells in the same notebook even if the entire code is written in Python. Getting error that unable to identify variable in the new cell
- 2882 Views
- 4 replies
- 1 kudos
- 1 kudos
Hi, thanks for your message. Can you sent a screenshot of your notebook maybe?This makes it hard to debug what could be the cause of it?Also which runtime are you using, this could might be the issue. With my experience loading the pandas library sho...
- 1 kudos
- 5308 Views
- 4 replies
- 0 kudos
Not able to log xgboost model to mlflow
I have been trying to log mlflow model but seems to be not working. It logs only the last(which is also the worst run).#-------------------------------------------------------13.0 ML XGBOost------------------------------------------------------------...
- 5308 Views
- 4 replies
- 0 kudos
- 0 kudos
@Kumaran Ran this code, but any specific log that I should be looking for?
- 0 kudos
- 1411 Views
- 1 replies
- 0 kudos
MlflowException: Unsupported Databricks profile key prefix: ''. Key prefixes cannot be empty.
I am trying to fetch data from mlflow model registry in Databricks and to use it in my local notebook. But I don't find any resource in internet to do so. I want to configure my mlflow in such a way i can fetch model registry values from databricks w...
- 1411 Views
- 1 replies
- 0 kudos
- 0 kudos
Hi @AnnamalaiVR,Thank you for posting the question in Databricks Community.In your Python code, import the MLflow library and create a client object to access your Model Registry. %pythonimport mlflow# Set the Databricks tracking URIdatabricks_host =...
- 0 kudos
- 7685 Views
- 2 replies
- 2 kudos
Resolved! Differences between Feature Store and Unity Catalog
Our small team has just finished the data preparation phase of our project and started data analysis in Databricks. As we go deeper into this field, we're trying to understand the distinctions and appropriate uses for a Feature Store versus a Unity C...
- 7685 Views
- 2 replies
- 2 kudos
- 2 kudos
Hi @Northp Good day!1.) A Feature Store is a centralized repository that enables data scientists to find and share features, ensuring that the same code used to compute the feature values is used for model training and inference. It is particularly...
- 2 kudos
- 4494 Views
- 4 replies
- 8 kudos
Azure - Databricks - account storage gen 2
Hello Every one, i am really new to databricks, just passed my apache developer certification on it.i also have a certification on data engineering with Azure.some fancy words here but i only started doing real deep work on them as i started a person...
- 4494 Views
- 4 replies
- 8 kudos
- 8 kudos
Hi,If we go by the error , Invalid configuration value detected for fs.azure.account.keyStorage account access key to access data using the abfssprotocol cannot be used. Please refer this https://learn.microsoft.com/en-us/azure/databricks/storage/azu...
- 8 kudos
- 4889 Views
- 6 replies
- 1 kudos
CUDA out of memory
I am trying out the new Meta LLama2 model.Following the databricks provided notebook example: https://github.com/databricks/databricks-ml-examples/blob/master/llm-models/llamav2/llamav2-13b/01_load_inference.py I keep getting CUDA out of memory. My G...
- 4889 Views
- 6 replies
- 1 kudos
- 1 kudos
Hi @Kumaran Hope you are well. Just wanted to see if you were able to find an answer to your question and would you like to mark an answer as best? It would be really helpful for the other members too. Cheers!
- 1 kudos
- 3458 Views
- 3 replies
- 0 kudos
- 3458 Views
- 3 replies
- 0 kudos
- 0 kudos
You can take all the Databricks exams as many times as you want, but you have to pay a fee each time you take the exam.
- 0 kudos
- 9526 Views
- 2 replies
- 1 kudos
Running test inference on Llama-2-70B-chat-GPTQ… are C++ libraries installed correctly?
Hi all,I was following the hugging face model https://huggingface.co/TheBloke/Llama-2-70B-chat-GPTQ, which points to use Exllama (https://github.com/turboderp/exllama/), which has 4 bit quantization.Running on a A10-Single-GPU-64GB,I've cloned the Ex...
- 9526 Views
- 2 replies
- 1 kudos
- 1 kudos
Hi @Kumaran,Thanks so much for the quick reply. When I run the script with !bash install_cusparse.shIt runs for a bit, but ultimately encounters an error. When I run !ls -l, i dont even see a data-mle directory in dbfshere is the full output from run...
- 1 kudos
- 1819 Views
- 2 replies
- 1 kudos
E-mail notification on failure run with DBX deployment
I am deploying workflow to Databricks using DBX. Here I want to add that when the workflow runs and if it fails I will get an e-mail on my_email@email.com. I have included an example workflow. deployments: - name: my_workflow ... # Other wo...
- 1819 Views
- 2 replies
- 1 kudos
- 1 kudos
Hi @akc,Thank you for posting your question in the Databricks community.Please refer to this documentation for the email notification.
- 1 kudos
- 6570 Views
- 1 replies
- 0 kudos
Cannot re-initialize CUDA in forked subprocess.
This is the error I am getting :"RuntimeError: Cannot re-initialize CUDA in forked subprocess. To use CUDA with multiprocessing, you must use the 'spawn' start method". I am using 13.0nc12s_v3 Cluster.I used this one :"import torch.multiprocessing as...
- 6570 Views
- 1 replies
- 0 kudos
- 0 kudos
Hi @phdykd,Thank you for posting your question in the Databricks community.One approach is to include the start_method="fork" parameter in the spawn function call as follows: mp.spawn(*prev_args, start_method="fork"). Although this will work, it migh...
- 0 kudos
- 1454 Views
- 0 replies
- 0 kudos
Unable to Infer Spark ML Pipeline model when built using Custom Preprocessing Stages
We are trying to build an internal use case based on PySpark. The data we have requires a lot of pre-processing. Hence, to cater to that we have used custom Spark ML pipeline stages as some of the transformations that need to be done on our data aren...
- 1454 Views
- 0 replies
- 0 kudos
- 2893 Views
- 4 replies
- 4 kudos
Resolved! Pyspark streaming optimization we need to focus on
What optimization we should focus on?
- 2893 Views
- 4 replies
- 4 kudos
- 4 kudos
@YanhDong_68817 This document is one of the good places to start evaluating our streaming pipeline - https://docs.databricks.com/structured-streaming/production.html
- 4 kudos
