With Effort you can adjust smoothly – and in real time – how many calculations you’d like to do during inference of an LLM model.

At 50% calculations it is as fast as regular matrix multiplications on Apple Silicon chips. At 25% effort it’s twice as fast and still retains most of the quality.

You can also freely choose to skip loading the least important weights.

It is implemented for Mistral now, it should work for all the other models just as well. No retraining needed, just conversion to a different format and some precomputation. — Read More

This work introduces an efficient method to scale Transformer-based Large Language Models (LLMs) to infinitely long inputs with bounded memory and computation. A key component in our proposed approach is a new attention technique dubbed Infini-attention. The Infini-attention incorporates a compressive memory into the vanilla attention mechanism and builds in both masked local attention and long-term linear attention mechanisms in a single Transformer block. We demonstrate the effectiveness of our approach on long-context language modeling benchmarks, 1M sequence length passkey context block retrieval and 500K length book summarization tasks with 1B and 8B LLMs. Our approach introduces minimal bounded memory parameters and enables fast streaming inference for LLMs. — Read More