AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration
Find the 1% of weights that matter by watching activations, then scale to protect them at INT4.
Lin et al. · MLSys 2024 · Serving. Read the paper ↗
A free, interactive, animated visual explainer of AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration — every exhibit computed from the real formulas, with verbatim quotes from the source.
Questions
- What is AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration?
- Find the 1% of weights that matter by watching activations, then scale to protect them at INT4.
- Who published AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration, and where?
- Lin et al. — MLSys 2024 (arXiv:2306.00978).
- Where can I find a visual explainer of AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration?
- Right here — a free, interactive, animated walkthrough of the whole paper, with exhibits computed from the real formulas and verbatim quotes from the source.
Related explainers
- PagedAttention (vLLM)
- Efficiently Scaling Transformer Inference
- Mooncake: A KVCache-centric Disaggregated Architecture for LLM Serving
- Fast Inference from Transformers via Speculative Decoding
- EAGLE: Speculative Sampling Requires Rethinking Feature Uncertainty
- DistServe: Disaggregating Prefill and Decoding for Goodput-optimized Large Language Model Serving
- CacheBlend: Fast Large Language Model Serving for RAG with Cached Knowledge Fusion
- Efficient Streaming Language Models with Attention Sinks