Sold by: nannyML
Deployed on AWS
Free Trial
Estimate the performance of your classification and regression models in production, without ground truth
Overview
In production ground truth is often delayed or absent. Traditional data drift detection techniques are noisy and do not only alert to changes that impact model performance.
Performance estimation allows you to estimate performance metrics (ROC-AUC, F1, RMSE, etc) without ground truth. Giving you a single metric to monitor, optimize and communicate about your models in production.
Some specific examples of when you could benefit from estimating your performance include:
When predicting loan defaults, to estimate model performance before the end of the repayment periods.
In demand forecasting, the ground truth demand will only be known after the forecast window has passed. Esimating performance lets you know how your model is performaning in real time.
When performing sentiment analysis, targets may be entirely unavailable without significant human effort, so estimation is the only feasible way to attain metrics.
Highlights
- Estimate the performance of machine learning models in production when targets are absent or delayed.
- NannyML supports Confidence Based Performance Estimation (CBPE) for performance estimation of binary and multiclass classification models.
- NannyML supports Direct Loss Estimation (DLE) for performance estimation of regression models.
Details
New
Introducing multi-product solutions
You can now purchase comprehensive solutions tailored to use cases and industries.
Features and programs
Financing for AWS Marketplace purchases
AWS Marketplace now accepts line of credit payments through the PNC Vendor Finance program. This program is available to select AWS customers in the US, excluding NV, NC, ND, TN, & VT.
Pricing
Free trial
Try this product free for 7 days according to the free trial terms set by the vendor.
Pricing is based on actual usage, with charges varying according to how much you consume. Subscriptions have no end date and may be canceled any time.
Additional AWS infrastructure costs may apply. Use the AWS Pricing Calculator to estimate your infrastructure costs.
Dimension | Description | Cost/host/hour |
|---|---|---|
ml.m5.large Inference (Batch) Recommended | Model inference on the ml.m5.large instance type, batch mode | $14.00 |
ml.m5.large Inference (Real-Time) Recommended | Model inference on the ml.m5.large instance type, real-time mode | $14.00 |
ml.m5.large Training Recommended | Algorithm training on the ml.m5.large instance type | $9.00 |
ml.m4.4xlarge Inference (Batch) | Model inference on the ml.m4.4xlarge instance type, batch mode | $14.00 |
ml.m5.4xlarge Inference (Batch) | Model inference on the ml.m5.4xlarge instance type, batch mode | $14.00 |
ml.m5.12xlarge Inference (Batch) | Model inference on the ml.m5.12xlarge instance type, batch mode | $14.00 |
ml.m4.16xlarge Inference (Batch) | Model inference on the ml.m4.16xlarge instance type, batch mode | $14.00 |
ml.m5.2xlarge Inference (Batch) | Model inference on the ml.m5.2xlarge instance type, batch mode | $14.00 |
ml.c4.4xlarge Inference (Batch) | Model inference on the ml.c4.4xlarge instance type, batch mode | $14.00 |
ml.m5.xlarge Inference (Batch) | Model inference on the ml.m5.xlarge instance type, batch mode | $14.00 |
Dimensions summary
You pay by the hour based on the compute instance you run. Charges cover three usage types: algorithm training, batch inference, and real-time inference. Each type is priced across a range of instance families and sizes, from the ml.m5.large up to the ml.m5.24xlarge. Larger instances hold more compute and memory. Your total cost depends on which instance you select, which usage type you run, and how many host-hours you consume. There is no fixed commitment; billing scales with the hours each instance runs.
Top-of-mind questions for buyers
A host-hour is one hour that a chosen compute instance runs. You pay for each instance while it is active. Counting starts when the instance launches for training or inference and stops when it shuts down. Larger instances hold more compute and memory, so their hourly rate is higher.
Training, batch inference, and real-time inference bill independently by host-hours. Each runs on its own instance and meters its own hours. Your invoice adds the host-hours from every usage type and instance you run. No single charge is bundled with another; each accrues separately based on running time.
Both meter host-hours on the instance you select. Batch inference runs predictions on stored datasets in scheduled jobs, so you pay only while the batch job runs. Real-time inference keeps an endpoint running to serve live requests, so charges accrue for as long as that endpoint stays active.
www.nannyml.com
Helpful?
Vendor refund policy
Given the free trial we do not support refunds, but you can cancel your subscription to the service at any time.
Legal
Vendor terms and conditions
Upon subscribing to this product, you must acknowledge and agree to the terms and conditions outlined in the vendor's End User License Agreement (EULA) .
Content disclaimer
Vendors are responsible for their product descriptions and other product content. AWS does not warrant that vendors' product descriptions or other product content are accurate, complete, reliable, current, or error-free.
Delivery details
Amazon SageMaker algorithm
An Amazon SageMaker algorithm is a machine learning model that requires your training data to make predictions. Use the included training algorithm to generate your unique model artifact. Then deploy the model on Amazon SageMaker for real-time inference or batch processing. Amazon SageMaker is a fully managed platform for building, training, and deploying machine learning models at scale.
Deploy the model on Amazon SageMaker AI using the following options:
Version release notes
Release Production Version!
Additional details
Inputs
- Summary
The input should be a CSV file. It should contain the names of the columns in the first row.
The required columns depend on the "parameters" defined during training. For more information read NannyML Performance Estimation Documentation .
The required number of rows depend on the chunking method defined during training.
- Limitations for input type
- The first line of the file should be the columns names, and it should contain the columns defined on the "parameters" during training. For realtime, the maximum size of the input data per invocation is 6 MB. For batch, the maximum size of the input data per invocation is 100 MB.
- Input MIME type
- text/csv
https://raw.githubusercontent.com/NannyML/sagemaker_perf_est_docs/main/notebooks/data/bc_reference.csv
https://raw.githubusercontent.com/NannyML/sagemaker_perf_est_docs/main/notebooks/data/bc_analysis.csv
Input data descriptions
The following table describes supported input data fields for real-time inference and batch transform.
Field name | Description | Constraints | Required |
|---|---|---|---|
y_pred | This column type is required on all machine learning problem types. The values are the predicted labels for classification and predicted number for regression.
Notice that the column name mapped to this column type on the "parameters" hyperparameter is defined during training, so you can have a different name for it on your CSV file. | Type: FreeText
Limitations: For classification data type can be text or integer but for regression it will be continuous. | Yes |
y_pred_proba | This column type is required on classification problem types. The values are the predicted scores or probabilities for a specific class.
Notice that the column name mapped to this column type on the "parameters" hyperparameter is defined during training, so you can have a different name for it on your CSV file. | Default value: No default values
Type: Continuous | No |
feature_column_names | This input refers to a list of required columns when using the performance estimation algorithm on a regression problem.
The list should include all the column names that should be consider features of the model whose performance we are predicting. | Default value: No default values
Type: FreeText
Limitations: The values are the features of your model. These can be categorical or continuous. NannyML identifies this based on their declared pandas data types. | No |
y_true | This column type contains actual model targets and is required on all machine learning problem types for the training data only.
Notice that the column name mapped to this column type on the "parameters" hyperparameter is defined during training, so you can have a different name for it on your CSV file.
[Optional] If you include this column on your inference data, realized performance will also be calculated separately to estimated performance to facilitate easy comparison. | Default value: No default values
Type: FreeText
Limitations: For classification data type can be text or integer but for regression it will be continuous. | No |
Resources
Support
Vendor support
If you have any questions, reach out to support@nannyml.com
AWS infrastructure support
AWS Support is a one-on-one, fast-response support channel that is staffed 24x7x365 with experienced and technical support engineers. The service helps customers of all sizes and technical abilities to successfully utilize the products and features provided by Amazon Web Services.
Similar products
Measure the impact of concept shift on your model's performance.
NannyML's Multicalibrated Confidence Based Performance Estimation Algorithm
Cohere provides a representative AI model: Embed that translates text into numerical vectors that models can understand.
Pre-trained Scikit Decision Trees model - Canonical Example
This model detects and counts how many people enter a building or a particular boundary line.
Text embedding model optimized for general-purpose (including multilingual) retrieval/search and AI applications. 32K context length.
Cohere provides a multilingual, representative AI model, that translates texts and images into numerical vectors that models can understand.
Customer reviews
No customer reviews yet
Be the first to review this product . We've partnered with PeerSpot to gather customer feedback. You can share your experience by writing or recording a review, or scheduling a call with a PeerSpot analyst.