[Doc] Sync four function docstrings with their signatures (#45437)

Signed-off-by: Daoyuan Li <94409450+DaoyuanLi2816@users.noreply.github.com>
Co-authored-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
This commit is contained in:
Daoyuan Li
2026-07-14 13:10:13 -04:00
committed by GitHub
co-authored by Wentao Ye mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
parent b2f7d2560a
commit ca3618bc69
3 changed files with 8 additions and 4 deletions
+5 -1
View File
@@ -1539,7 +1539,11 @@ def scaled_fp4_quant(
Args:
input: The input tensor to be quantized to FP4
input_global_scale: A scalar scaling factor for the entire tensor.
use_8x4_sf_layout: Whether to use the 8x4 or 128x4 layout for the scaling
is_sf_swizzled_layout: Whether to store the scaling factors in the
swizzled layout (default `True`).
backend: Quantization kernel backend to dispatch to. For `"trtllm"`
backends the 8x4 scale-factor layout is selected for small
batches (m <= 32) instead of the 128x4 layout.
padded_n: Optional padded K dimension. When provided, the quantized
output and scale tensors are allocated for ``padded_n``