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

Tiny cleanup the models' name in test_utils (#15920)

Browse Source
This commit is contained in:
Liangsheng Yin
2025-12-27 14:13:23 +08:00
committed by GitHub
parent 29ce7b3612
commit 9ad546d7e8
15 changed files with 88 additions and 87 deletions
Download Patch File Download Diff File Expand all files Collapse all files
test/srt/hicache/test_hicache_variants.py
+4 -4
View File
@@ -12,10 +12,10 @@ from sglang.bench_serving import get_tokenizer
from sglang.srt.utils import is_hip, kill_process_tree
from sglang.test.run_eval import run_eval
from sglang.test.test_utils import (
DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST_EAGLE3,
DEFAULT_DRAFT_MODEL_EAGLE3,
DEFAULT_MLA_MODEL_NAME_FOR_TEST,
DEFAULT_MODEL_NAME_FOR_TEST,
DEFAULT_MODEL_NAME_FOR_TEST_EAGLE3,
DEFAULT_TARGET_MODEL_EAGLE3,
DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
DEFAULT_URL_FOR_TEST,
CustomTestCase,
@@ -113,7 +113,7 @@ class TestHiCacheMLA(HiCacheBaseServer, HiCacheEvalMixin, HiCacheMGSMEvalMixin):
class TestHiCacheEagle(HiCacheBaseServer, HiCacheEvalMixin):
"""HiCache with EAGLE speculative decoding tests"""
model_name = DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST_EAGLE3
model_name = DEFAULT_TARGET_MODEL_EAGLE3
needs_tokenizer = True
hicache_args = [
"--enable-hierarchical-cache",
@@ -124,7 +124,7 @@ class TestHiCacheEagle(HiCacheBaseServer, HiCacheEvalMixin):
"--speculative-algorithm",
"EAGLE3",
"--speculative-draft-model-path",
DEFAULT_MODEL_NAME_FOR_TEST_EAGLE3,
DEFAULT_DRAFT_MODEL_EAGLE3,
"--speculative-num-steps",
2,
"--speculative-eagle-topk",
test/srt/openai_server/features/test_openai_server_hidden_states.py
+6 -6
View File
@@ -6,9 +6,9 @@ import openai
from sglang.srt.utils import kill_process_tree
from sglang.srt.utils.hf_transformers_utils import get_tokenizer
from sglang.test.test_utils import (
DEFAULT_EAGLE_DRAFT_MODEL_FOR_TEST,
DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST,
DEFAULT_DRAFT_MODEL_EAGLE,
DEFAULT_SMALL_MODEL_NAME_FOR_TEST,
DEFAULT_TARGET_MODEL_EAGLE,
DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
DEFAULT_URL_FOR_TEST,
CustomTestCase,
@@ -259,10 +259,10 @@ class TestOpenAIServerWithEAGLEAndHiddenStatesEnabled(
):
@classmethod
def setUpClass(cls):
cls.model = DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST
cls.model = DEFAULT_TARGET_MODEL_EAGLE
cls.base_url = DEFAULT_URL_FOR_TEST
cls.api_key = "sk-123456"
cls.speculative_draft_model = DEFAULT_EAGLE_DRAFT_MODEL_FOR_TEST
cls.speculative_draft_model = DEFAULT_DRAFT_MODEL_EAGLE
cls.speculative_algorithm = "EAGLE"
cls.process = popen_launch_server(
cls.model,
@@ -272,7 +272,7 @@ class TestOpenAIServerWithEAGLEAndHiddenStatesEnabled(
"--speculative-algorithm",
"EAGLE",
"--speculative-draft-model-path",
DEFAULT_EAGLE_DRAFT_MODEL_FOR_TEST,
DEFAULT_DRAFT_MODEL_EAGLE,
"--speculative-num-steps",
5,
"--speculative-eagle-topk",
@@ -289,7 +289,7 @@ class TestOpenAIServerWithEAGLEAndHiddenStatesEnabled(
],
)
cls.base_url += "/v1"
cls.tokenizer = get_tokenizer(DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST)
cls.tokenizer = get_tokenizer(DEFAULT_TARGET_MODEL_EAGLE)
cls.return_hidden_states = [False, True]
cls.use_list_input = [True, False]
cls.parallel_sample_nums = [1]
test/srt/test_bench_serving.py
+4 -4
View File
@@ -5,14 +5,14 @@ import unittest
import requests
from sglang.test.test_utils import (
DEFAULT_EAGLE_DRAFT_MODEL_FOR_TEST,
DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST,
DEFAULT_DRAFT_MODEL_EAGLE,
DEFAULT_MODEL_NAME_FOR_TEST,
DEFAULT_MODEL_NAME_FOR_TEST_FP8,
DEFAULT_MOE_MODEL_NAME_FOR_TEST,
DEFAULT_SMALL_EMBEDDING_MODEL_NAME_FOR_TEST,
DEFAULT_SMALL_MODEL_NAME_FOR_TEST_SCORE,
DEFAULT_SMALL_VLM_MODEL_NAME_FOR_TEST,
DEFAULT_TARGET_MODEL_EAGLE,
CustomTestCase,
is_in_amd_ci,
is_in_ci,
@@ -328,7 +328,7 @@ class TestBenchServing(CustomTestCase):
def test_online_latency_eagle(self):
res = run_bench_serving(
model=DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST,
model=DEFAULT_TARGET_MODEL_EAGLE,
num_prompts=300,
request_rate=8,
sharegpt_context_len=3072,
@@ -338,7 +338,7 @@ class TestBenchServing(CustomTestCase):
"--speculative-algorithm",
"EAGLE",
"--speculative-draft-model-path",
DEFAULT_EAGLE_DRAFT_MODEL_FOR_TEST,
DEFAULT_DRAFT_MODEL_EAGLE,
"--speculative-num-steps",
"5",
"--speculative-eagle-topk",
test/srt/test_disaggregation_basic.py
+4 -4
View File
@@ -12,9 +12,9 @@ from sglang.test.server_fixtures.disaggregation_fixture import (
PDDisaggregationServerBase,
)
from sglang.test.test_utils import (
DEFAULT_EAGLE_DRAFT_MODEL_FOR_TEST,
DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST,
DEFAULT_DRAFT_MODEL_EAGLE,
DEFAULT_MODEL_NAME_FOR_TEST,
DEFAULT_TARGET_MODEL_EAGLE,
DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
popen_launch_pd_server,
)
@@ -280,8 +280,8 @@ class TestDisaggregationMooncakeSpec(PDDisaggregationServerBase):
@classmethod
def setUpClass(cls):
super().setUpClass()
cls.model = DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST
cls.draft_model = DEFAULT_EAGLE_DRAFT_MODEL_FOR_TEST
cls.model = DEFAULT_TARGET_MODEL_EAGLE
cls.draft_model = DEFAULT_DRAFT_MODEL_EAGLE
cls.spec_args = [
"--speculative-algorithm",
"EAGLE",
test/srt/test_eagle_dp_attention.py
+4 -4
View File
@@ -6,8 +6,8 @@ import requests
from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
from sglang.test.send_one import BenchArgs, send_one_prompt
from sglang.test.test_utils import (
DEFAULT_EAGLE_DP_ATTENTION_DRAFT_MODEL_FOR_TEST,
DEFAULT_EAGLE_DP_ATTENTION_TARGET_MODEL_FOR_TEST,
DEFAULT_DRAFT_MODEL_EAGLE_DP_ATTN,
DEFAULT_TARGET_MODEL_EAGLE_DP_ATTN,
DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
DEFAULT_URL_FOR_TEST,
CustomTestCase,
@@ -22,7 +22,7 @@ from sglang.test.test_utils import (
class TestEAGLE3EngineDPAttention(CustomTestCase):
@classmethod
def setUpClass(cls):
cls.model = DEFAULT_EAGLE_DP_ATTENTION_TARGET_MODEL_FOR_TEST
cls.model = DEFAULT_TARGET_MODEL_EAGLE_DP_ATTN
cls.base_url = DEFAULT_URL_FOR_TEST
other_args = [
"--trust-remote-code",
@@ -35,7 +35,7 @@ class TestEAGLE3EngineDPAttention(CustomTestCase):
"--speculative-num-draft-tokens",
"32",
"--speculative-draft-model-path",
DEFAULT_EAGLE_DP_ATTENTION_DRAFT_MODEL_FOR_TEST,
DEFAULT_DRAFT_MODEL_EAGLE_DP_ATTN,
"--tp-size",
"2",
"--dp-size",
test/srt/test_standalone_speculative_decoding.py
+5 -5
View File
@@ -7,8 +7,8 @@ from sglang.srt.environ import envs
from sglang.srt.utils import kill_process_tree
from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
from sglang.test.test_utils import (
DEFAULT_STANDALONE_SPECULATIVE_DRAFT_MODEL_FOR_TEST,
DEFAULT_STANDALONE_SPECULATIVE_TARGET_MODEL_FOR_TEST,
DEFAULT_DRAFT_MODEL_STANDALONE,
DEFAULT_TARGET_MODEL_STANDALONE,
DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
DEFAULT_URL_FOR_TEST,
CustomTestCase,
@@ -26,7 +26,7 @@ DEFAULT_SERVER_ARGS = [
"--speculative-algorithm",
"STANDALONE",
"--speculative-draft-model-path",
DEFAULT_STANDALONE_SPECULATIVE_DRAFT_MODEL_FOR_TEST,
DEFAULT_DRAFT_MODEL_STANDALONE,
"--speculative-num-steps",
"4",
"--speculative-eagle-topk",
@@ -40,8 +40,8 @@ DEFAULT_SERVER_ARGS = [
class TestStandaloneSpeculativeDecodingBase(CustomTestCase):
model = DEFAULT_STANDALONE_SPECULATIVE_TARGET_MODEL_FOR_TEST
draft_model = DEFAULT_STANDALONE_SPECULATIVE_DRAFT_MODEL_FOR_TEST
model = DEFAULT_TARGET_MODEL_STANDALONE
draft_model = DEFAULT_DRAFT_MODEL_STANDALONE
base_url = DEFAULT_URL_FOR_TEST
accuracy_threshold = 0.7 # derived tests need to override this
spec_decode_threshold = 3.6 # derived spec decoding tests need to override this