Mitigating the Impact of Outlier Channels for Language Model Quantization with Activation Regularization

arXiv:2404.03605v1 Announce Type: new
Abstract: We consider the problem of accurate quantization for language models, where both the weights and activations are uniformly quantized to 4 bits per parameter, the lowest bitwidth format natively supported by GPU hardware. In this context, the key challenge is activation quantization: it is known that language models contain outlier channels whose values on average are orders of magnitude higher than than other channels, which prevents accurate low-bitwidth quantization with known techniques. We systematically study …

abstract arxiv challenge channels context cs.cl cs.lg format gpu hardware impact key language language model language models outlier per quantization regularization the key type