TildAlice – Tech Blog for Developers
Python, AI, Deep Learning, Data Analysis, Computer Vision, Reinforcement Learning โ tips and tutorials
Latest Posts
-
ONNX Export Pitfalls: 7 PyTorch โ Production Gotchas
Fix ONNX export failures fast. Learn the 7 critical PyTorch-to-production gotchas that silently break modelsโand how to catch them before deployment.
-
FlashAttention-2 Warmup: Fix 3x Slower First Batch
First FlashAttention-2 batch is 3x slower? Fix kernel compilation overhead with warmup, persistent cache, and bucketingโreal latency numbers included.
-
Monte Carlo VaR Underestimates Tail Risk: 3 Distribution Fixes
Monte Carlo VaR fails in extreme markets. Compare Gaussian, t-distribution, and GARCH models to capture fat tails and improve risk estimates.
-
GitHub Actions ML Pipeline: First CI/CD Portfolio Project
Build an automated ML training pipeline with GitHub Actions, MLflow tracking, and live metrics dashboard โ proof you ship production-ready code.
-
PPO vs SAC: Real Robot Benchmark on 3 Manipulation Tasks
Compare PPO vs SAC on real robot manipulation tasks. Performance metrics, training stability, and practical deployment insights revealed.
-
Docker vs Poetry vs uv: 3 Setup Patterns That Actually Scale
Compare Docker, Poetry, and uv for Python project setup. See which pattern saves time, avoids dependency hell, and scales to production.
-
Ruff vs Black vs Flake8: 1000-File Accuracy Benchmark
Ruff caught 847 issues in 1000 files. Flake8 found 634 but took 18x longer. See which bugs each tool missed and why speed isn't the only metric.
-
Feature Stores Are Overengineered: When SQL Is Enough
Most teams waste months on Feast or Tecton when Postgres + cron does the job. Here's when SQL beats feature stores โ with real latency numbers.