Accelerate deepseek fp4 b200 ci (#11993)
This commit is contained in:
@@ -32,6 +32,8 @@ class TestDeepseekV3FP4(CustomTestCase):
|
|||||||
"flashinfer_trtllm",
|
"flashinfer_trtllm",
|
||||||
"--quantization",
|
"--quantization",
|
||||||
"modelopt_fp4",
|
"modelopt_fp4",
|
||||||
|
"--kv-cache-dtype",
|
||||||
|
"fp8_e4m3",
|
||||||
]
|
]
|
||||||
cls.process = popen_launch_server(
|
cls.process = popen_launch_server(
|
||||||
cls.model,
|
cls.model,
|
||||||
@@ -100,6 +102,9 @@ class TestDeepseekV3FP4MTP(CustomTestCase):
|
|||||||
"1",
|
"1",
|
||||||
"--speculative-num-draft-tokens",
|
"--speculative-num-draft-tokens",
|
||||||
"4",
|
"4",
|
||||||
|
"--kv-cache-dtype",
|
||||||
|
"fp8_e4m3",
|
||||||
|
"--enable-beta-spec",
|
||||||
]
|
]
|
||||||
cls.process = popen_launch_server(
|
cls.process = popen_launch_server(
|
||||||
cls.model,
|
cls.model,
|
||||||
|
|||||||
Reference in New Issue
Block a user