Practical guides for ML practitioners and data engineers.
No fluff. Tutorials that go from problem to working code. Error fixes with exact tracebacks and root causes.
Browse articlesDeveloper toolsWork with meFix a common error
- GPU / CUDA Out of Memory & Device Errors
- Hugging Face Model Loading & Caching Errors
- scikit-learn Errors
- pandas Data-Wrangling Errors
- LLM Pipelines, Embeddings & Retrieval
- Python ML Environment & Import Errors
Google DeepMind's Gemini Robotics 2 is a three-model system (a VLA, an embodied-reasoning model, and an on-device VLA) that controls a whole humanoid body, not only the upper body used for table-top tasks in earlier models.
Fix scikit-learn's ValueError: could not convert string to float, caused by passing categorical/string columns straight into a model. Covers get_dummies, LabelEncoder pitfalls, and a proper ColumnTransformer pipeline.
Fix scikit-learn's UserWarning: X does not have valid feature names / X has feature names, but..., caused by mixing DataFrames and NumPy arrays across fit/transform. Includes the ValueError variant.
Fix XGBoost ValueError: DataFrame.dtypes for data must be int, float, bool or category. Reproduced on XGBoost 3.4.1 with pandas 3 and 2.2: find the column, then cast, encode or use category dtype.
Fix ValueError: numpy.dtype size changed, may indicate binary incompatibility. Reproduced with pandas, scikit-learn, scipy and OpenCV wheels built for NumPy 1.x running on NumPy 2, with the fixes that worked.
Fix cannot import name 'is_offline_mode', 'cached_download' or 'HfFolder' from 'huggingface_hub'. Reproduced on real version combinations, with the pins and code changes that fixed each.
Measured on faiss 1.15.1: nprobe above nlist adds nothing, -1 results from small nprobe, the exact IndexIVFPQ d % M and nbits errors, bytes per vector, and GPU alloc fail.
Fix 'Asking to pad but the tokenizer does not have a padding token', 'Unable to create tensor', ignored max_length and right-padding warnings in transformers. Reproduced on transformers 5.17.0.
Fix 'No module named langchain_openai', old langchain.chat_models imports, 'Input should be an instance of Runnable', ServiceContext and embedding dimension errors. Reproduced on langchain 1.4.2.
A practical guide to data drift alerting in production ML systems: what to monitor, how to query model outputs against ground truth in ClickHouse, when to alert, and how to support rollback decisions.
Free, in-browser tools for ML engineers, data scientists, and developers.
GPU VRAM EstimatorEstimate GPU memory for inference or training from parameter count and precision.Token CounterCount tokens for GPT, Claude, and other LLMs before you hit an API context limit.Confusion Matrix CalculatorPrecision, recall, F1, and accuracy from a confusion matrix, for classification model evaluation.Neural Network Parameter CounterStack PyTorch-style layers and see total parameters and memory footprint.Regex TesterTest regular expressions against sample text with live match highlighting.View all tools →