CERWU adds a quadratic rate estimate to the layer-wise loss and uses Optimal Brain Surgeon updates to produce quantized weights that entropy-code 20-40% smaller than NNCodec at equal accuracy on CNNs.
EfficientQAT: Efficient Quantization-Aware Training for Large Language Models, October 2024
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Reducing Storage of Pretrained Neural Networks by Rate-Constrained Quantization and Entropy Coding
CERWU adds a quadratic rate estimate to the layer-wise loss and uses Optimal Brain Surgeon updates to produce quantized weights that entropy-code 20-40% smaller than NNCodec at equal accuracy on CNNs.