Commit ff39c79
committed
Add WaterSIC KV-cache quantization plot example
Self-contained script that reproduces the KV-cache quantization plots
comparing RTN, GPTQ, and WaterSIC methods across temperature, power-law,
retrieval, sink, and conditioning sweeps. Imports WaterSIC from the
ported zsic module and inlines baseline quantization utilities.
Signed-off-by: Kai Xu <kaix@nvidia.com>1 parent 3807e6d commit ff39c79
1 file changed
+1176
-0
lines changed
0 commit comments