EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

Fix retract for page size > 1 (#4914)

Browse Source
This commit is contained in:
Lianmin Zheng
2025-03-30 02:57:15 -07:00
committed by GitHub
parent b26bc86b36
commit 4ede6770cd
10 changed files with 68 additions and 120 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
test/srt/models/lora/test_lora_tp.py
View File

@@ -33,6 +33,9 @@ CI_LORA_MODELS = [
],
max_loras_per_batch=1,
),
]
ALL_OTHER_LORA_MODELS = [
LoRAModelCase(
base="meta-llama/Llama-3.1-8B-Instruct",
adaptors=[
@@ -43,9 +46,6 @@ CI_LORA_MODELS = [
],
max_loras_per_batch=1,
),
]
ALL_OTHER_LORA_MODELS = [
LoRAModelCase(
base="meta-llama/Llama-2-7b-hf",
adaptors=[LoRAAdaptor(name="winddude/wizardLM-LlaMA-LoRA-7B")],

16
test/srt/run_suite.py
View File

@@ -16,7 +16,7 @@ suites = {
TestFile("models/lora/test_lora.py", 76),
TestFile("models/lora/test_lora_backend.py", 420),
TestFile("models/lora/test_multi_lora_backend.py", 144),
TestFile("models/test_embedding_models.py", 119),
TestFile("models/test_embedding_models.py", 35),
TestFile("models/test_generation_models.py", 103),
TestFile("models/test_grok_models.py", 60),
TestFile("models/test_qwen_models.py", 82),
@@ -38,7 +38,7 @@ suites = {
TestFile("test_metrics.py", 32),
TestFile("test_mla.py", 92),
TestFile("test_mla_deepseek_v3.py", 221),
TestFile("test_mla_int8_deepseek_v3.py", 421),
TestFile("test_mla_int8_deepseek_v3.py", 522),
TestFile("test_mla_flashinfer.py", 395),
TestFile("test_mla_fp8.py", 93),
TestFile("test_no_chunked_prefill.py", 126),
@@ -59,7 +59,7 @@ suites = {
TestFile("test_srt_endpoint.py", 94),
TestFile("test_torch_compile.py", 76),
TestFile("test_torch_compile_moe.py", 85),
TestFile("test_torch_native_attention_backend.py", 149),
TestFile("test_torch_native_attention_backend.py", 123),
TestFile("test_torchao.py", 70),
TestFile("test_triton_attention_kernels.py", 4),
TestFile("test_triton_attention_backend.py", 134),
@@ -76,6 +76,16 @@ suites = {
TestFile("test_hicache.py", 60),
TestFile("test_hicache_mla.py", 90),
],
"per-commit-2-gpu": [
TestFile("test_data_parallelism.py", 90),
TestFile("test_dp_attention.py", 90),
TestFile("test_update_weights_from_distributed.py", 100),
TestFile("test_verl_engine.py", 100),
TestFile("test_patch_torch.py", 30),
TestFile("test_moe_ep.py", 220),
TestFile("test_mla_tp.py", 420),
TestFile("test_lora_tp.py", 300),
],
"nightly": [
TestFile("test_nightly_gsm8k_eval.py"),
],

4
test/srt/test_dp_attention.py
View File

@@ -60,3 +60,7 @@ class TestDPAttentionDP2TP2(CustomTestCase):
metrics = run_eval(args)
print(f"{metrics=}")
self.assertGreater(metrics["score"], 0.8)
if __name__ == "__main__":
unittest.main()

1
test/srt/test_metrics.py
View File

@@ -63,7 +63,6 @@ class TestEnableMetrics(CustomTestCase):
"sglang:cached_tokens_total",
"sglang:num_requests_total",
"sglang:time_to_first_token_seconds",
"sglang:time_per_output_token_seconds",
"sglang:inter_token_latency_seconds",
"sglang:e2e_request_latency_seconds",
]