TildAlice – Tech Blog for Developers
Python, AI, Deep Learning, Data Analysis, Computer Vision, Reinforcement Learning โ tips and tutorials
Latest Posts
-
Gold Price Patterns: 10 Years of Market Data Analysis
10 years of gold price data reveals 3 volatility regimes and non-stationary patterns that ARIMA and LSTM models systematically miss.
-
Chinchilla Review: Why Fewer Parameters Beat Gopher
Chinchilla's 70B model beat 280B Gopher using 4x fewer parameters. The secret: training tokens matter more than model size. Proof inside.
-
DVC vs MLflow: Data Versioning and Model Registry Compared
DVC tracks data pipelines with Git commits. MLflow logs experiments and serves models. Compare both with real codeโchoose the right tool.
-
Python contextlib: contextmanager and ExitStack Deep Dive
Stop writing repetitive try-finally blocks. Python's @contextmanager and ExitStack manage resources in 3 linesโbenchmarks and gotchas included.
-
RL Transfer Learning: Atari to Custom Tasks with SB3
Transfer a pretrained DQN from Atari Pong to custom tasks in 100 episodes. Replay buffer tricks and exploration tuning that actually work.
-
๐โโ๏ธ Today’s Journal: Keeping the Pace
Building a daily running habit after weight gain: minimum time targets, pace strategies, and staying consistent without burning out.
-
Whisper Architecture: How OpenAI’s Speech Model Works
Whisper's encoder-decoder uses 1.5GB VRAM for Large model. Break down the architecture and memory bottlenecks before mobile deployment.
-
SimCLR for Bearing Fault Classification with Few Labels
SimCLR pretraining hits 89.3% accuracy with only 5 labeled bearing fault samples. Step-by-step contrastive learning on CWRU dataset.