Focused Transformer: Contrastive Training for Context Scaling (256k context length) #1135

Focused Transformer: Contrastive Training for Context Scaling (256k context length) #1135

Source