5 papers
Mirror Speculative Decoding: Breaking the Serial Barrier in LLM Inference
Nikhil Bhendawade, Kumari Nishu, Arnav Kundu +3
Speculative decoding accelerates LLM inference by using a draft model to look ahead, but gains are capped by the cost of autoregressive draft generation: increasing draft size elev…
MoEs Are Stronger than You Think: Hyper-Parallel Inference Scaling with RoE
Soheil Zibakhsh, Mohammad Samragh, Kumari Nishu +3
The generation quality of large language models (LLMs) is often improved by utilizing inference-time sequence-level scaling methods (e.g., Chain-of-Thought). We introduce hyper-par…
MoE-PHDS: One MoE checkpoint for flexible runtime sparsity
Lauren. A Hannah, Soheil Zibakhsh, Kumari Nishu +4
Sparse Mixtures of Experts (MoEs) are typically trained to operate at a fixed sparsity level, e.g. in a top- gating function. This global sparsity level determines an operat…
Your LLM Knows the Future: Uncovering Its Multi-Token Prediction Potential
Mohammad Samragh, Arnav Kundu, David Harrison +4
Autoregressive language models are constrained by their inherently sequential nature, generating one token at a time. This paradigm limits inference speed and parallelism, especial…
From Dense to Dynamic: Token-Difficulty Driven MoEfication of Pre-Trained LLMs
Kumari Nishu, Sachin Mehta, Samira Abnar +6
Training large language models (LLMs) for different inference constraints is computationally expensive, limiting control over efficiency-accuracy trade-offs. Moreover, once trained…