Model metadata and lineage in MLOps - Step-by-Step Execution

Process Flow - Model metadata and lineage

Start: Model Created

↓

Capture Metadata

↓

Store Metadata

↓

Track Lineage

↓

Update Metadata & Lineage on Changes

↓

Use Metadata & Lineage for Audits/Debug

↓

End

The flow shows how model metadata is captured, stored, and lineage tracked through changes to support audits and debugging.

Execution Sample

MLOps

model = create_model('v1')
metadata = capture_metadata(model)
store_metadata(metadata)
lineage = track_lineage(model)
update_metadata(metadata, 'v2')
model = create_model('v2')
lineage = track_lineage(model)

This code creates a model, captures and stores its metadata, tracks lineage, then updates metadata for a new version and tracks lineage for the new model.

Process Table

Step	Action	Input	Output	System State Change
1	Create model	'v1'	Model object v1	Model instance created
2	Capture metadata	Model object v1	Metadata {version: 'v1', created_at: timestamp}	Metadata extracted from model
3	Store metadata	Metadata {version: 'v1', created_at: timestamp}	Stored in metadata DB	Metadata saved in database
4	Track lineage	Model object v1	Lineage record {parent: null, model: 'v1'}	Lineage entry created
5	Update metadata	Metadata, new version 'v2'	Metadata updated {version: 'v2', updated_at: timestamp}	Metadata record updated
6	Create model	'v2'	Model object v2	Model instance created
7	Track lineage	Model object v2	Lineage record {parent: 'v1', model: 'v2'}	Lineage updated with new version
8	Use metadata & lineage	Query metadata DB	Audit report generated	Audit completed
9	End	-	-	Process complete

💡 Process ends after metadata and lineage are updated and used for audit.

Status Tracker

Variable	Start	After Step 1	After Step 2	After Step 5	After Step 7	Final
model	null	Model v1 object	Model v1 object	Model v1 object	Model v2 object	Model v2 object
metadata	null	null	{version: 'v1', created_at: timestamp}	{version: 'v2', updated_at: timestamp}	{version: 'v2', updated_at: timestamp}	{version: 'v2', updated_at: timestamp}
lineage	null	null	Lineage {parent: null, model: 'v1'}	Lineage {parent: null, model: 'v1'}	Lineage {parent: 'v1', model: 'v2'}	Lineage {parent: 'v1', model: 'v2'}

Key Moments - 3 Insights

Why do we track lineage after updating the model version?

Because lineage shows the relationship between model versions, tracking it after updates (see steps 5 and 7) helps understand model evolution.

What happens if metadata is not stored after capture?

If metadata is not stored (step 3), it cannot be used later for audits or debugging, breaking traceability.

How does metadata help in audits?

Metadata contains details like version and timestamps (step 8), which auditors use to verify model history and changes.

Visual Quiz - 3 Questions

Look at the execution table, what is the metadata version after step 5?

A 'v1'

B null

C 'v2'

D 'v3'

💡 Hint

Check the Output column at step 5 in the execution table.

At which step is the lineage first created?

A Step 2

B Step 4

C Step 7

D Step 3

💡 Hint

Look for 'Lineage entry created' in the System State Change column.

If we skip storing metadata at step 3, what is the likely impact?

A Audit reports cannot be generated properly

B Lineage tracking will fail

C Model creation will fail

D Metadata will update automatically

💡 Hint

Refer to the key moment about metadata storage and audit usage.

Concept Snapshot

Model metadata records details like version and timestamps.
Lineage tracks model relationships and changes over time.
Capture metadata when creating or updating models.
Store metadata and lineage in a database.
Use metadata and lineage for audits and debugging.
Keep lineage updated with each model version.

Full Transcript

This visual execution shows how model metadata and lineage are managed in MLOps. First, a model is created, then metadata is captured and stored. Lineage is tracked to record relationships between model versions. When the model updates, metadata and lineage are updated accordingly. Finally, metadata and lineage are used for audits and debugging. Variables like model, metadata, and lineage change step-by-step, showing how the system state evolves. Key moments clarify why lineage is tracked after updates, why storing metadata is essential, and how metadata supports audits. The quiz tests understanding of metadata versions, lineage creation steps, and the impact of skipping metadata storage.