[router][ci] Quick Improvement to make CI more stable (#12869)
This commit is contained in:
@@ -25,11 +25,11 @@ from util import kill_process_tree
|
|||||||
|
|
||||||
|
|
||||||
class TestGrpcBackend(StateManagementTests, MCPTests, StructuredOutputBaseTest):
|
class TestGrpcBackend(StateManagementTests, MCPTests, StructuredOutputBaseTest):
|
||||||
"""End to end tests for gRPC backend (Regular backend with Llama)."""
|
"""End to end tests for gRPC backend (Regular backend with Qwen2.5)."""
|
||||||
|
|
||||||
@classmethod
|
@classmethod
|
||||||
def setUpClass(cls):
|
def setUpClass(cls):
|
||||||
cls.model = "/home/ubuntu/models/meta-llama/Llama-3.1-8B-Instruct"
|
cls.model = "/home/ubuntu/models/Qwen/Qwen2.5-14B-Instruct"
|
||||||
cls.base_url_port = "http://127.0.0.1:30030"
|
cls.base_url_port = "http://127.0.0.1:30030"
|
||||||
|
|
||||||
cls.cluster = popen_launch_workers_and_router(
|
cls.cluster = popen_launch_workers_and_router(
|
||||||
@@ -46,7 +46,7 @@ class TestGrpcBackend(StateManagementTests, MCPTests, StructuredOutputBaseTest):
|
|||||||
"--history-backend",
|
"--history-backend",
|
||||||
"memory",
|
"memory",
|
||||||
"--tool-call-parser",
|
"--tool-call-parser",
|
||||||
"llama",
|
"qwen",
|
||||||
],
|
],
|
||||||
)
|
)
|
||||||
|
|
||||||
@@ -152,7 +152,7 @@ class TestGrpcHarmonyBackend(
|
|||||||
cls.base_url_port,
|
cls.base_url_port,
|
||||||
timeout=90,
|
timeout=90,
|
||||||
num_workers=1,
|
num_workers=1,
|
||||||
tp_size=2,
|
tp_size=4,
|
||||||
policy="round_robin",
|
policy="round_robin",
|
||||||
worker_args=[
|
worker_args=[
|
||||||
"--reasoning-parser=gpt-oss",
|
"--reasoning-parser=gpt-oss",
|
||||||
|
|||||||
@@ -86,10 +86,16 @@ class FunctionCallingBaseTest(ResponseAPIBaseTest):
|
|||||||
},
|
},
|
||||||
},
|
},
|
||||||
]
|
]
|
||||||
|
system_prompt = (
|
||||||
|
"You are a helpful assistant that can call functions. "
|
||||||
|
"When a user asks for horoscope information, call the function. "
|
||||||
|
"IMPORTANT: Don't reply directly to the user, only call the function. "
|
||||||
|
)
|
||||||
|
|
||||||
# Create a running input list we will add to over time
|
# Create a running input list we will add to over time
|
||||||
input_list = [
|
input_list = [
|
||||||
{"role": "user", "content": "What is my horoscope? I am an Aquarius."}
|
{"role": "system", "content": system_prompt},
|
||||||
|
{"role": "user", "content": "What is my horoscope? I am an Aquarius."},
|
||||||
]
|
]
|
||||||
|
|
||||||
# 2. Prompt the model with tools defined
|
# 2. Prompt the model with tools defined
|
||||||
|
|||||||
Reference in New Issue
Block a user