Tools, Techniques and Best Practices

Explain the rationale and aim
Give examples of some tools and applications
Radius-graph/knn from FAISS
Edge matching
Truth graph builder (Pandas building, would be nice to get this on CuPy)
Triplet selection (possibly use the PyEmbedding version - benchmark)
Triplet/line graph converter
Checkpointing
Mixed precision
Remove duplicate edges
Cell feature calculation
Training best practices (data balancing, hard negative mining, load balancing)