chenchenghao/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

[NVIDIA] Add fp8 gemm benchmark on blackwell (#13528)

Browse Source
This commit is contained in:
Kaixi Hou
2025-11-19 19:35:00 -08:00
committed by GitHub
parent dc69462456
commit c3c4da71fb
2 changed files with 331 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
python/sglang/srt/layers/quantization/fp8_utils.py
View File

@@ -439,7 +439,7 @@ def requant_weight_ue8m0_inplace(weight, weight_scale_inv, weight_block_size):
assert isinstance(weight, torch.nn.Parameter)
assert isinstance(weight_scale_inv, torch.nn.Parameter)
new_weight, new_weight_scale_inv = _requant_weight_ue8m0(
new_weight, new_weight_scale_inv = requant_weight_ue8m0(
weight.to(weight_scale_inv.device), weight_scale_inv, weight_block_size
)
@@ -447,7 +447,7 @@ def requant_weight_ue8m0_inplace(weight, weight_scale_inv, weight_block_size):
weight_scale_inv.data = new_weight_scale_inv
def _requant_weight_ue8m0(
def requant_weight_ue8m0(
weight: torch.Tensor,
weight_scale_inv: torch.Tensor,
weight_block_size: List[int],