4-bit and 6-bit integer-only quantized Transformers implemented on Spartan-7 FPGA for AIoT time-series forecasting achieve 0.63% higher test loss than 8-bit baselines but up to 132x speedup and 48x lower energy.
I-ViT: integer-only quantization for efficient vision Transformer inference
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2024 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Integer-only Quantized Transformers for Embedded FPGA-based Time-series Forecasting in AIoT
4-bit and 6-bit integer-only quantized Transformers implemented on Spartan-7 FPGA for AIoT time-series forecasting achieve 0.63% higher test loss than 8-bit baselines but up to 132x speedup and 48x lower energy.