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/registered/spec/eagle/test_eagle_constrained_decoding.py
+4 -4
View File
@@ -6,8 +6,8 @@ from sglang.test.ci.ci_register import register_cuda_ci
from sglang.test.kits.json_constrained_kit import TestJSONConstrainedMixin
from sglang.test.kits.regex_constrained_kit import TestRegexConstrainedMixin
from sglang.test.test_utils import (
DEFAULT_EAGLE_DRAFT_MODEL_FOR_TEST,
DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST,
DEFAULT_DRAFT_MODEL_EAGLE,
DEFAULT_TARGET_MODEL_EAGLE,
DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
DEFAULT_URL_FOR_TEST,
CustomTestCase,
@@ -27,8 +27,8 @@ class TestEagleConstrainedDecoding(
spec_draft_tokens = 6
page_size = 1
other_launch_args = []
model = DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST
draft_model = DEFAULT_EAGLE_DRAFT_MODEL_FOR_TEST
model = DEFAULT_TARGET_MODEL_EAGLE
draft_model = DEFAULT_DRAFT_MODEL_EAGLE
grammar_backend = "xgrammar"
eagle_v2 = False
test/registered/spec/eagle/test_eagle_infer_a.py
+17 -17
View File
@@ -10,11 +10,11 @@ from sglang.srt.utils import kill_process_tree
from sglang.srt.utils.hf_transformers_utils import get_tokenizer
from sglang.test.ci.ci_register import register_cuda_ci
from sglang.test.test_utils import (
DEFAULT_EAGLE_DRAFT_MODEL_FOR_TEST,
DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST,
DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST_EAGLE3,
DEFAULT_MODEL_NAME_FOR_TEST_EAGLE3,
DEFAULT_DRAFT_MODEL_EAGLE,
DEFAULT_DRAFT_MODEL_EAGLE3,
DEFAULT_MODEL_NAME_FOR_TEST_MLA,
DEFAULT_TARGET_MODEL_EAGLE,
DEFAULT_TARGET_MODEL_EAGLE3,
DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
DEFAULT_URL_FOR_TEST,
CustomTestCase,
@@ -31,8 +31,8 @@ decode_tolerance: float = 5e-2
class TestEAGLEEngine(CustomTestCase):
BASE_CONFIG = {
"model_path": DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST,
"speculative_draft_model_path": DEFAULT_EAGLE_DRAFT_MODEL_FOR_TEST,
"model_path": DEFAULT_TARGET_MODEL_EAGLE,
"speculative_draft_model_path": DEFAULT_DRAFT_MODEL_EAGLE,
"speculative_algorithm": "EAGLE",
"speculative_num_steps": 5,
"speculative_eagle_topk": 4,
@@ -133,7 +133,7 @@ class TestEAGLEEngine(CustomTestCase):
"skip_special_tokens": False,
}
tokenizer = get_tokenizer(DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST)
tokenizer = get_tokenizer(DEFAULT_TARGET_MODEL_EAGLE)
output = engine.generate(prompt, params)["text"]
print(f"{output=}")
@@ -187,8 +187,8 @@ class TestEAGLEEngineTokenMap(TestEAGLEEngine):
class TestEAGLE3Engine(TestEAGLEEngine):
BASE_CONFIG = {
"model_path": DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST_EAGLE3,
"speculative_draft_model_path": DEFAULT_MODEL_NAME_FOR_TEST_EAGLE3,
"model_path": DEFAULT_TARGET_MODEL_EAGLE3,
"speculative_draft_model_path": DEFAULT_DRAFT_MODEL_EAGLE3,
"speculative_algorithm": "EAGLE3",
"speculative_num_steps": 5,
"speculative_eagle_topk": 16,
@@ -206,8 +206,8 @@ class TestEAGLE3Engine(TestEAGLEEngine):
class TestEAGLERadixCache(CustomTestCase):
BASE_CONFIG = {
"model_path": DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST_EAGLE3,
"speculative_draft_model_path": DEFAULT_MODEL_NAME_FOR_TEST_EAGLE3,
"model_path": DEFAULT_TARGET_MODEL_EAGLE3,
"speculative_draft_model_path": DEFAULT_DRAFT_MODEL_EAGLE3,
"speculative_algorithm": "EAGLE3",
"speculative_num_steps": 2,
"speculative_eagle_topk": 2,
@@ -309,14 +309,14 @@ class TestEAGLEDraftExtend(CustomTestCase):
def setUpClass(cls):
cls.base_url = DEFAULT_URL_FOR_TEST
cls.process = popen_launch_server(
DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST,
DEFAULT_TARGET_MODEL_EAGLE,
cls.base_url,
timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
other_args=[
"--speculative-algorithm",
"EAGLE",
"--speculative-draft-model-path",
DEFAULT_EAGLE_DRAFT_MODEL_FOR_TEST,
DEFAULT_DRAFT_MODEL_EAGLE,
"--speculative-num-steps",
1,
"--speculative-eagle-topk",
@@ -375,14 +375,14 @@ class TestEAGLEDraftExtendFlashinfer(TestEAGLEDraftExtend):
def setUpClass(cls):
cls.base_url = DEFAULT_URL_FOR_TEST
cls.process = popen_launch_server(
DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST,
DEFAULT_TARGET_MODEL_EAGLE,
cls.base_url,
timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
other_args=[
"--speculative-algorithm",
"EAGLE",
"--speculative-draft-model-path",
DEFAULT_EAGLE_DRAFT_MODEL_FOR_TEST,
DEFAULT_DRAFT_MODEL_EAGLE,
"--speculative-num-steps",
1,
"--speculative-eagle-topk",
@@ -404,14 +404,14 @@ class TestEAGLEDraftExtendTriton(TestEAGLEDraftExtend):
def setUpClass(cls):
cls.base_url = DEFAULT_URL_FOR_TEST
cls.process = popen_launch_server(
DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST,
DEFAULT_TARGET_MODEL_EAGLE,
cls.base_url,
timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
other_args=[
"--speculative-algorithm",
"EAGLE",
"--speculative-draft-model-path",
DEFAULT_EAGLE_DRAFT_MODEL_FOR_TEST,
DEFAULT_DRAFT_MODEL_EAGLE,
"--speculative-num-steps",
1,
"--speculative-eagle-topk",
test/registered/spec/eagle/test_eagle_infer_b.py
+2 -5
View File
@@ -15,10 +15,7 @@ from sglang.test.ci.ci_register import register_cuda_ci
from sglang.test.few_shot_gsm8k import run_eval as run_gsm8k_eval
from sglang.test.kits.radix_cache_server_kit import run_radix_attention_test
from sglang.test.server_fixtures.eagle_fixture import EagleServerBase
from sglang.test.test_utils import (
DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST,
run_logprob_check,
)
from sglang.test.test_utils import DEFAULT_TARGET_MODEL_EAGLE, run_logprob_check
register_cuda_ci(est_time=1100, suite="stage-b-test-small-1-gpu")
@@ -262,7 +259,7 @@ class TestEAGLEServerBasic(EagleServerBase):
response = requests.post(
self.base_url + "/v1/chat/completions",
json={
"model": DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST,
"model": DEFAULT_TARGET_MODEL_EAGLE,
"messages": messages,
"temperature": 0,
"response_format": {"type": "json_object"},
test/registered/spec/eagle/test_eagle_infer_beta.py
+4 -4
View File
@@ -8,8 +8,8 @@ from sglang.test.few_shot_gsm8k import run_eval
from sglang.test.kits.matched_stop_kit import MatchedStopMixin
from sglang.test.kits.radix_cache_server_kit import run_radix_attention_test
from sglang.test.test_utils import (
DEFAULT_EAGLE_DRAFT_MODEL_FOR_TEST,
DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST,
DEFAULT_DRAFT_MODEL_EAGLE,
DEFAULT_TARGET_MODEL_EAGLE,
DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
DEFAULT_URL_FOR_TEST,
CustomTestCase,
@@ -27,8 +27,8 @@ class TestEagleServerBase(CustomTestCase, MatchedStopMixin):
spec_draft_tokens = 6
page_size = 1
other_launch_args = []
model = DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST
draft_model = DEFAULT_EAGLE_DRAFT_MODEL_FOR_TEST
model = DEFAULT_TARGET_MODEL_EAGLE
draft_model = DEFAULT_DRAFT_MODEL_EAGLE
@classmethod
def setUpClass(cls):
test/registered/spec/test_ngram_speculative_decoding.py
+2 -2
View File
@@ -5,7 +5,7 @@ from sglang.srt.utils import kill_process_tree
from sglang.test.ci.ci_register import register_cuda_ci
from sglang.test.kits.gsm8k_accuracy_kit import GSM8KMixin
from sglang.test.test_utils import (
DEFAULT_NGRAM_SPECULATIVE_TARGET_MODEL_FOR_TEST,
DEFAULT_TARGET_MODEL_NGRAM,
DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
DEFAULT_URL_FOR_TEST,
CustomTestCase,
@@ -32,7 +32,7 @@ DEFAULT_SERVER_ARGS = [
class TestNgramSpeculativeDecodingBase(GSM8KMixin, CustomTestCase):
model = DEFAULT_NGRAM_SPECULATIVE_TARGET_MODEL_FOR_TEST
model = DEFAULT_TARGET_MODEL_NGRAM
base_url = DEFAULT_URL_FOR_TEST
gsm8k_accuracy_thres = 0.79 # derived tests need to override this
gsm8k_accept_length_thres = 1.8 # derived spec decoding tests need to override this