Ml/pr3 solution articles

[ahmadbasyouni10]

• File(s) Modified: articles/rms-normalization.md, articles/training-diagnostics.md
• Language(s) Used: python
• Submission URL: N/A (ML course solution articles, not LeetCode submissions)

Summary

Adds solution articles for 2 new ML problems:

RMS Normalization (rms-normalization.md):

• Explains RMSNorm vs LayerNorm (no mean subtraction, no beta parameter)
• Step-by-step formula walkthrough: RMS computation, normalize, scale by gamma
• Common pitfalls: forgetting epsilon, using beta parameter, wrong axis
• Project context: model/rms_normalization.py — used by modern LLMs (LLaMA, Mistral)

Training Diagnostics (training-diagnostics.md):

• Three diagnostic methods: compute_activation_stats, compute_gradient_stats, diagnose
• Diagnosis priority: dead_neurons → exploding_gradients → vanishing_gradients → healthy
• Thresholds and reasoning for each diagnosis category
• Project context: foundations/training_diagnostics.py — debug tool for the GPT training pipeline