chenchenghao/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 11 Packages Projects Releases Wiki Activity

[V32/GLM5] Control the threshold of applying dense attention with an environ (#20062)

Browse Source
This commit is contained in:
Baizhou Zhang
2026-03-09 14:36:10 -07:00
committed by GitHub
parent d39ed074cf
commit be63f982b7
6 changed files with 32 additions and 59 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
test/registered/quant/test_deepseek_v32_fp4_4gpu.py
View File

@@ -34,8 +34,6 @@ class TestDeepseekV32FP4DP(CustomTestCase):
"flashinfer_trtllm",
"--quantization",
"modelopt_fp4",
"--kv-cache-dtype",
"fp8_e4m3",
"--tool-call-parser",
"deepseekv32",
"--reasoning-parser",
@@ -103,8 +101,6 @@ class TestDeepseekV32FP4TP(CustomTestCase):
"flashinfer_trtllm",
"--quantization",
"modelopt_fp4",
"--kv-cache-dtype",
"fp8_e4m3",
"--tool-call-parser",
"deepseekv32",
"--reasoning-parser",

4
test/registered/quant/test_deepseek_v32_fp4_mtp_4gpu.py
View File

@@ -39,8 +39,6 @@ class TestDeepseekV32FP4DPSpecV2(CustomTestCase):
"flashinfer_trtllm",
"--quantization",
"modelopt_fp4",
"--kv-cache-dtype",
"fp8_e4m3",
"--tool-call-parser",
"deepseekv32",
"--reasoning-parser",
@@ -131,8 +129,6 @@ class TestDeepseekV32FP4TPSpecV2(CustomTestCase):
"flashinfer_trtllm",
"--quantization",
"modelopt_fp4",
"--kv-cache-dtype",
"fp8_e4m3",
"--tool-call-parser",
"deepseekv32",
"--reasoning-parser",