Hardware and framework version tracking in MLOps - Step-by-Step Execution

Process Flow - Hardware and framework version tracking

Start Training Job

↓

Detect Hardware Specs

↓

Log Hardware Info

↓

Detect Framework Version

↓

Log Framework Version

↓

Run Training

↓

Save Logs & Metadata

↓

End Training Job

The process starts by detecting hardware and framework versions, logging them, then running the training job while saving all metadata for reproducibility.

Execution Sample

MLOps

import torch
import platform

def log_versions():
    hw = platform.platform()
    fw = torch.__version__
    print(f"Hardware: {hw}")
    print(f"Framework: PyTorch {fw}")

This code detects and prints the hardware platform and PyTorch framework version.

Process Table

Step	Action	Detected Hardware	Detected Framework Version	Output
1	Call platform.platform()	Linux-5.15.0-1051-azure-x86_64-with-glibc2.29		Hardware info string returned
2	Access torch.__version__		2.0.1+cu117	Framework version string returned
3	Print hardware info	Linux-5.15.0-1051-azure-x86_64-with-glibc2.29		Hardware: Linux-5.15.0-1051-azure-x86_64-with-glibc2.29
4	Print framework version		2.0.1+cu117	Framework: PyTorch 2.0.1+cu117
5	End of function	Linux-5.15.0-1051-azure-x86_64-with-glibc2.29	2.0.1+cu117	Versions logged successfully

💡 All hardware and framework versions detected and logged, function completes.

Status Tracker

Variable	Start	After Step 1	After Step 2	Final
hw	None	Linux-5.15.0-1051-azure-x86_64-with-glibc2.29	Linux-5.15.0-1051-azure-x86_64-with-glibc2.29	Linux-5.15.0-1051-azure-x86_64-with-glibc2.29
fw	None	None	2.0.1+cu117	2.0.1+cu117

Key Moments - 2 Insights

Why do we detect hardware info before framework version?

Detecting hardware first ensures we know the environment before checking software versions, as shown in execution_table steps 1 and 2.

What if the framework version is not detected correctly?

If framework version is missing, logging will be incomplete, which can cause reproducibility issues; see step 2 where fw is assigned.

Visual Quiz - 3 Questions

Look at the execution table at step 3, what is printed for hardware info?

A Hardware: Linux-5.15.0-1051-azure-x86_64-with-glibc2.29

B Framework: PyTorch 2.0.1+cu117

C Hardware info string returned

D Versions logged successfully

💡 Hint

Check the Output column at step 3 in the execution_table.

At which step is the framework version detected?

A Step 1

B Step 4

C Step 2

D Step 5

💡 Hint

Look at the Action and Detected Framework Version columns in execution_table.

If hardware detection failed at step 1, what would happen to variable 'hw'?

A 'hw' would have an error string

B 'hw' would remain None

C 'hw' would be set to framework version

D 'hw' would be empty string

💡 Hint

Refer to variable_tracker for 'hw' initial and after step 1 values.

Concept Snapshot

Hardware and framework version tracking:
- Detect hardware info (e.g., platform.platform())
- Detect framework version (e.g., torch.__version__)
- Log both before running training
- Essential for reproducibility and debugging
- Save metadata with training outputs

Full Transcript

This visual execution shows how hardware and framework versions are detected and logged in an MLOps context. First, the system queries the hardware platform using platform.platform(), storing the result in variable 'hw'. Next, it accesses the framework version from torch.__version__, storing it in 'fw'. Both values are printed to the console to confirm detection. The execution table traces each step, showing the values assigned and printed. The variable tracker highlights how 'hw' and 'fw' change from None to their detected strings. Key moments clarify why hardware is detected before framework and the importance of successful detection for reproducibility. The quiz tests understanding of the printed outputs, detection steps, and variable states. This process ensures that training jobs record their environment details, helping teams reproduce results and debug issues effectively.