Thea:

• FastML workshop 2025

Andrew:

• No raw data, only aggregation: best results! 8*16+8*16(8 sets, 16 features sums and means)
• Sparse attention mechanism: pre-train constituent BDT, add score per constituent
• Finish document on data preparation
• Thea send latency estimate script

Patrick:

• Chat with Anant on KD --> Patrick 
•