Medusa: Simple LLM Inference Acceleration Framework with Multiple Decoding Heads
Paper
•
2401.10774
•
Published
•
54
fast
-
-
SliceGPT: Compress Large Language Models by Deleting Rows and Columns
Paper • 2401.15024 • Published • 69 -
Speculative Streaming: Fast LLM Inference without Auxiliary Models
Paper • 2402.11131 • Published • 42 -
SLAB: Efficient Transformers with Simplified Linear Attention and Progressive Re-parameterized Batch Normalization
Paper • 2405.11582 • Published • 13