Mitigating the Impact of Outlier Channels for Language Model Quantization with Activation Regularization
Name
nrusimha-anin-sm-eecs-2024-thesis.pdf
Description
Thesis PDF
Size
3.77 MB
Format
Adobe PDF
Checksum (MD5)
0d0ab50f468370a5aab271414cc4a9fb
Author(s)
Nrusimha, Aniruddha
Advisor(s)
Kim, Yoon
Date Issued
May 2024
Publisher
Massachusetts Institute of Technology
Abstract
We consider the problem of accurate quantization for language models, where both the weights and activations are quantized to 4 bits per parameter with uniform quantization, the lowest bitwidth format natively supported by existing 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 this phenomena and find that these outlier channels emerge early in training, and that they occur more frequently in layers with residual streams. We then propose a simple strategy which regularizes a layer’s inputs via quantization-aware training (QAT) and its outputs via activation kurtosis regularization. We show that regularizing both the inputs and outputs is crucial for preventing a model’s "migrating" the difficulty in input quantization to the weights, which makes post-training quantization (PTQ) of weights more difficult. When combined with weight PTQ, we show that our approach can obtain a W4A4 model with integer quantization that performs competitively to the standard-precision W16A16 baseline.1
MIT Department
Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science
Terms of Use
Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0)
Copyright retained by author(s)
Persistent DSpace Link
