Model stages (staging, production, archived) in MLOps - Step-by-Step Execution

Process Flow - Model stages (staging, production, archived)

Model Created

↓

Stage: Staging

↓

Test & Validate

↓

Promote

↓

Stage: Production

↓

Monitor & Use

↓

Archive Old Models

Models start in staging for testing, then move to production if validated, and old models get archived.

Execution Sample

MLOps

model = create_model()
stage = 'staging'
if test_model(model):
  stage = 'production'
else:
  stage = 'staging'

This code creates a model, tests it, and sets its stage to production if tests pass, otherwise keeps it in staging.

Process Table

Step	Action	Model Stage	Test Result	Next Stage
1	Create model	staging	N/A	staging
2	Test model	staging	Pass	production
3	Promote model	production	N/A	production
4	Monitor model	production	N/A	production
5	Archive old model	production	N/A	archived

💡 Model promoted to production after passing tests, then archived when replaced.

Status Tracker

Variable	Start	After Step 1	After Step 2	After Step 3	After Step 5
stage	None	staging	production	production	archived
test_result	None	None	Pass	None	None

Key Moments - 2 Insights

Why does the model stay in staging if tests fail?

Because the test result is 'No' (see concept_flow ASCII diagram), the model is not promoted and remains in staging for fixes.

What triggers moving a model to archived?

When a newer model replaces the production one, the old model stage changes to archived (see execution_table step 5).

Visual Quiz - 3 Questions

Look at the execution_table at step 2, what is the model stage after testing?

A production

B staging

C archived

D testing

💡 Hint

Check the 'Next Stage' column at step 2 in execution_table.

At which step does the model get archived?

A Step 3

B Step 4

C Step 5

D Step 2

💡 Hint

Look for 'archived' in the 'Next Stage' column in execution_table.

If the test_model function fails, what will be the model stage after step 2?

A archived

B staging

C production

D deleted

💡 Hint

Refer to variable_tracker and execution_table step 2 for test results and stage changes.

Concept Snapshot

Model stages track lifecycle:
- Staging: model testing and validation
- Production: live use after passing tests
- Archived: old models replaced by newer ones
Models move forward only if tests pass.

Full Transcript

In MLOps, models go through stages to manage their lifecycle. First, a model is created and placed in the staging stage where it is tested. If the tests pass, the model is promoted to production, meaning it is ready for real use. If tests fail, the model stays in staging for fixes and retesting. When a newer model replaces an old one, the old model is moved to the archived stage to keep history and avoid confusion. This flow ensures only good models serve users and old versions are safely stored.