transformers
7 posts tagged transformers.
- What LoRA Rank Actually Buys You, According to the Attention Math
- RecipeNet Treats Procedures as Structure, Not Flat Tables
- Sparse Weight Decomposition makes circuit extraction less expensive
- Depth breaks rank before it breaks loss
- Transformer circuits may be lower-dimensional than they look
- PEHT makes traffic forecasting less about bigger Transformers
- Tapered Language Models and the Free Lever in Layer Width