How to check whether all types of ML training were completed successfully in [ml_solution] including Predictive Intelligence Solutions, NLU Models, NLU Batch Tests, Vocabulary Sources, Process Optimization, Document Intelligence and other AI capabilitiesSummaryAll AI/ML capabilities on the ServiceNow Platform have their solutions trained via table [ml_solution]. 1. Open the List View for table [ml_solution] You should add the Capability and Updated fields to the List View on table [ml_solution], as a minimum and if you sort by the Updated field, if you have just submitted a ML Solution for training, you should see the record in with the state Waiting for Training and when the training is successful, it will change the state to Solution Complete. However, on occasions, the solution training may fail, which will be indicated via the State field on the [ml_solution] record. ML Solution Training states: Here is the list of the different Solution training states. If the state on the [ml_solution] record is - Solution Complete:It means the solution training was successful and it will have generated the [ml_model_artifact] records, which will have 1 or more records in table [ml_model_artifact] for each successfully trained [ml_solution] with a [sys_attachment] on each [ml_model_artifact] record. Configuration OR Network Error:This indicates there is an issue with the ML configuration. Please review Configuration tips for Predictive Intelligence. Please engage ServiceNow Technical Support, if this issue persists. Error while training solution:This indicates there is an issue with the ML Solution Definition and will need to be reviewed. Please engage ServiceNow Technical Support to check the error and provide the solution. In most cases, the issue is with the solution definition itself. We will be improving our error messaging, so that when the solution training fails, it will provide an informative error message with possible steps to resolution. Waiting for Training:Some capabilities only take a minute or two to train, such as NLU Model training, but other capabilities can take 12 hours or more to complete depending on the size of the training dataset, and we support up to 300k records. If the datacenter has many ML training requests in a short timespan, a backlog can occur and it can wait for up to 4 hours, before it changes the state to Training request timed out. It will automatically generate a new [ml_solution] record for a second attempt and third attempt until such a time, where if it is still unsuccessful on the last attempt, it will change the state to Training is Cancelled on the last [ml_solution] record that was created for this ML Solution training. Please engage ServiceNow Technical Support, if this issue persists, but usually once the backlog has cleared, the solution training will eventually be successful. Fetching Files for Training:You may get an error when in the state Fetching Files for Training, as the [sharedservice.worker] user does not have access to the training dataset. To resolve this issue, please follow the Resolution section in KB0826655: [ML] Predictive Intelligence solution training fails at SE0060:Solution Training Failed. Ask Support to use log key XXXXXX to investigate trainer logs further., logging details: java.lang.Exception: SE0089:Failed to download dataset(s) from glide.null or engage ServiceNow Technical Support, if you need further assistance ML Capabilities: Here is the list of machine learning capabilities on the ServiceNow Platform that use solution training for machine learning. More capabilities will be added in the future. Note: This ML Capabilities list needs updating. Agent Zero [agent_zero_trainer]Proactively deflect issues and expedite the resolution process with Issue Auto Resolution. Provide users with immediate self-service by using machine learning to intelligently deliver Virtual Agent topics, Knowledge articles, and catalog items. Classification [classification_trainer]The Predictive Intelligence classification framework enables you to use machine-learning algorithms to set field values during record creation, such as setting the incident category based on the short description. You can train predictive models so they act as an agent to automatically categorize and route work based on your past record-handling experience. Clustering [clustering_trainer]The Predictive Intelligence clustering framework enables you to group similar records into clusters, so you can address them collectively or identify patterns. For example, you can group similar incidents that have occurred recently to identify a major incident. Fuzzy Matcher [fuzzy_matcher_trainer]Synchronize your Vocabulary Sources for your NLU Models to obtain the latest changes to the ServiceNow source table. Synchronizing your vocabulary sources ensures your models have the latest values when predicting intents. Genius Results Assembly [genius_results_assembly]Display the best answers for a search query as actionable Genius Result cards included with search results used by the AI Search engine NLU [nlu_trainer]Use NLU models to apply ServiceNow Natural Language Understanding on your instances. Create, manage, train, test, and publish NLU models with the NLU Workbench.Multi-model Batch Testing - Test multiple Natural Language Understanding (NLU) models against a large set of utterances to evaluate the performance of the models. Process Optimization (Mining) [promin_trainer]Process Optimization helps analysts and process owners quickly analyze and optimize their business processes on the ServiceNow Platform. Process Optimization (Cluster Analysis) [promin_show_record_trainer]When identifying an activity, connection, or route as a potential bottleneck, view clusters of keyword descriptions and assignment groups to gain insights. Regression [regression_trainer]The Predictive Intelligence regression framework is a machine-learning framework that you can train with historic data to predict numeric outputs, such as a temperature or a stock price. For example, you can use regression to estimate the time it takes to resolve an incident or a case. Search Relevancy [search_relevancy_trainer]AI Search displays the most relevant search results for a query first. Machine learning automatically tunes search result relevancy scoring for search experiences based on aggregated user interactions. Note: Machine learning relevancy is automatically enabled and isn't configurable. Similarity [similarity_trainer]The Predictive Intelligence similarity framework identifies existing records that have similar values to a new record. You use the framework to build a word corpus. The word corpus functions as the vocabulary the system uses to compare your trained records based on their textual similarity. For example, you can train a subset of your incident records to recommend a resolution based on the information of a similar incident record. By reusing similar closed incidents that have a proven resolution, you can help agents and fulfillers quickly provide the best resolution for an incoming incident. Workflow [workflow_trainer]Automated root cause analysis - Find where and why inefficiencies occur within your processes using automated root cause analysis.Automation Discovery - Helps you identify automation opportunities for your workflows.Document Intelligence - An artificial intelligence (AI) solution that enables any organization to automate and accelerate the process of extracting data from documents. That data can easily be integrated into larger automation workflows to save time and resources. Source of Truth: The instance is the persistent layer and source of truth for all solution training on all AI capabilities within the ServiceNow Platform. All solutions are trained via table [ml_solution], and once the training has successfully completed, it will upload the trained solution to the instance and generate records in table [ml_model_artifact] where each record contains an attachment [sys_attachment] of the trained solution. The AI capability can have one or more records in table [ml_model_artifact] associated with it. Here are the steps to find the associated trained solution records - 1. Open table [ml_solution] and find the solution based on the Solution Name, Capability and Updated fields. Right-click on the record and click on "Copy sys_id". 2. Open table [ml_model_artifact] and click on the Filter icon and select "Show Related Fields" in the -- choose field -- drop-down list field. 3. Select the [sys_id] field on the related Solution table "Solution.Sys ID", set the operator to "is" and paste the [sys_id] copied in Step 1 into the -- value -- field. 4. Click on the "Run" button and it will return all the [ml_model_artifact] records, usually between 1 and 6 records. Each [ml_model_artifact] record will contain an attachment with the trained solution. You can download the attachment and open it in a Text Editor, as some of these attachments are in readable format, but others will be in a compiled format. If the [ml_model_artifact] records are missing their attachments, please review KB0861056: Clone does not preserve ML Solutions when setting property.Related Links1. Transfer a trained Predictive Intelligence solution to another instance using the following steps - KB0997816: How to move your Predictive Intelligence Model to another instance with the Trained Solution Note: You can also train the Predictive Intelligence solution on the target instance that will recreate the [ml_model_artifact] records and the attachments containing the trained solution. 2. Using the attachment from ML Model Store for a trained NLU Model, you can easily compare NLU Models between instances using the following steps - KB0994548: How to compare and check the differences of your NLU Model on two different instances