Skip to content

[Feature] A calibration-free RTN-based quantization for accurate and accelerated INT4/INT8 inference #8009

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
wants to merge 7 commits into
base: main
Choose a base branch
from
+759 −0
Draft

[Feature] A calibration-free RTN-based quantization for accurate and accelerated INT4/INT8 inference #8009

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
39a690d
add rtn quant
YouNeedCryDear Jul 13, 2025
27ea1e1
enable rtn in cmd arg and init
YouNeedCryDear Jul 13, 2025
c024e8f
add rtn moe support
YouNeedCryDear Jul 14, 2025
5d0f7b7
add test cases for rtn quant
YouNeedCryDear Jul 15, 2025
a59516c
fix test case
YouNeedCryDear Jul 15, 2025
5523e48
add rtn utils
YouNeedCryDear Jul 15, 2025
66a5afd
fix moe bug for rtn
YouNeedCryDear Jul 16, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions python/sglang/srt/layers/quantization/__init__.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -68,6 +68,7 @@ def override_quantization_method(self, *args, **kwargs):
)
from sglang.srt.layers.quantization.moe_wna16 import MoeWNA16Config
from sglang.srt.layers.quantization.qoq import QoQConfig
from sglang.srt.layers.quantization.rtn import RTNConfig
from sglang.srt.layers.quantization.w4afp8 import W4AFp8Config
from sglang.srt.layers.quantization.w8a8_fp8 import W8A8Fp8Config
from sglang.srt.layers.quantization.w8a8_int8 import W8A8Int8Config
Expand All @@ -84,6 +85,7 @@ def override_quantization_method(self, *args, **kwargs):
"compressed-tensors": CompressedTensorsConfig,
"qoq": QoQConfig,
"w4afp8": W4AFp8Config,
"rtn": RTNConfig,
}

# VLLM-dependent quantization methods
Expand Down
Loading