SpinQuant learns optimal rotations to enable accurate 4-bit quantization of LLM weights, activations, and KV cache, reducing the zero-shot gap to full precision to 2.9 points on LLaMA-2 7B.
Outlier suppression: Pushing the limit of low-bit transformer language models, 2022a
4 Pith papers cite this work. Polarity classification is still indexing.
4
Pith papers citing it
representative citing papers
LLM.int8() performs 8-bit inference for transformers up to 175B parameters with no accuracy loss by combining vector-wise quantization for most features with 16-bit mixed-precision handling of systematic outlier dimensions.
AWQ quantizes LLM weights to low bits by scaling salient channels based on activation statistics, outperforming prior methods on language, coding, math, and multi-modal benchmarks.
LoRaQ enables fully sub-16-bit quantized diffusion models by optimizing low-rank error compensation in a data-free way, outperforming prior methods at equal memory cost on Pixart-Σ and SANA while supporting mixed low-precision branches.
citing papers explorer
-
SpinQuant: LLM quantization with learned rotations
SpinQuant learns optimal rotations to enable accurate 4-bit quantization of LLM weights, activations, and KV cache, reducing the zero-shot gap to full precision to 2.9 points on LLaMA-2 7B.
-
LLM.int8(): 8-bit Matrix Multiplication for Transformers at Scale
LLM.int8() performs 8-bit inference for transformers up to 175B parameters with no accuracy loss by combining vector-wise quantization for most features with 16-bit mixed-precision handling of systematic outlier dimensions.
-
AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration
AWQ quantizes LLM weights to low bits by scaling salient channels based on activation statistics, outperforming prior methods on language, coding, math, and multi-modal benchmarks.
-
LoRaQ: Optimized Low Rank Approximation for 4-bit Quantization
LoRaQ enables fully sub-16-bit quantized diffusion models by optimizing low-rank error compensation in a data-free way, outperforming prior methods at equal memory cost on Pixart-Σ and SANA while supporting mixed low-precision branches.