minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 13 Packages Projects Releases Wiki Activity

[NVIDIA] [3/N] Nvfp4 Masked Gemm: Add flashinfer grouped_gemm_nt_masked (#9199)

Browse Source
This commit is contained in:
Shu Wang
2025-09-11 20:18:43 -07:00
committed by GitHub
parent 7b141f816c
commit 3df05f4d6a
11 changed files with 694 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/sglang/test/test_utils.py
+3
View File
@@ -53,6 +53,9 @@ DEFAULT_MLA_FP8_MODEL_NAME_FOR_TEST = "neuralmagic/DeepSeek-Coder-V2-Lite-Instru
DEFAULT_MODEL_NAME_FOR_TEST_MLA = "lmsys/sglang-ci-dsv3-test"
DEFAULT_MODEL_NAME_FOR_TEST_MLA_NEXTN = "lmsys/sglang-ci-dsv3-test-NextN"
# NVFP4 models
DEFAULT_DEEPSEEK_NVFP4_MODEL_FOR_TEST = "nvidia/DeepSeek-R1-0528-FP4"
# FP8 models
DEFAULT_MODEL_NAME_FOR_TEST_FP8 = "neuralmagic/Meta-Llama-3.1-8B-Instruct-FP8"
DEFAULT_MODEL_NAME_FOR_ACCURACY_TEST_FP8 = "neuralmagic/Meta-Llama-3.1-8B-Instruct-FP8"