EngineX/xc-llm-ascend
3
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

[CI] Drop ascend scheduler from test (#4613)

Browse Source 
Drop ascend scheduler from test

- vLLM version: v0.11.2

Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
This commit is contained in:
wangxiyuan
2025-12-02 13:18:17 +08:00
committed by GitHub
parent 6360eb1dea
commit 400af665e6
7 changed files with 1 additions and 165 deletions
Download Patch File Download Diff File Expand all files Collapse all files

118
tests/e2e/singlecard/test_ascend_scheduler.py
View File

@@ -1,118 +0,0 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import pytest
from vllm import SamplingParams
from tests.e2e.conftest import VllmRunner
from tests.e2e.model_utils import check_outputs_equal
MODEL = "Qwen/Qwen3-0.6B"
@pytest.mark.parametrize("enforce_eager", [True, False])
def test_concurrent_partial_prefill(enforce_eager):
with VllmRunner(MODEL,
max_num_seqs=3,
max_num_batched_tokens=8192,
enforce_eager=enforce_eager,
gpu_memory_utilization=0.7) as vllm_model:
outputs = vllm_model.model.generate(["Hello my name is Robert and I"] *
3)
assert len(outputs) == 3
for output in outputs:
assert len(output.outputs) == 1
@pytest.mark.parametrize("enforce_eager", [True, False])
def test_prefix_cache_stats_is_recorded(enforce_eager):
with VllmRunner(MODEL,
max_num_seqs=3,
max_num_batched_tokens=8192,
enforce_eager=enforce_eager,
gpu_memory_utilization=0.7) as vllm_model:
# 17 tokens will make sure first 16 tokens are cached in a block
input_tokens = {"prompt_token_ids": [101] * 129}
_ = vllm_model.model.generate([input_tokens])
outputs = vllm_model.model.generate([input_tokens])
assert outputs[0].num_cached_tokens == 128
@pytest.mark.parametrize("max_tokens",
[4]) # cannot align results when max_tokens > 4
@pytest.mark.parametrize("chunked_prefill_token_size", [2048])
def test_chunked_prefill_with_scheduler_dynamic_batch(
max_tokens: int, chunked_prefill_token_size: int) -> None:
example_prompts = [
"vLLM is a high-throughput and memory-efficient inference and serving engine for LLMs."
]
max_num_seqs = chunked_prefill_token_size
max_num_batched_tokens = chunked_prefill_token_size
with VllmRunner(MODEL,
additional_config={
'SLO_limits_for_dynamic_batch': 0,
},
max_num_seqs=max_num_seqs,
max_num_batched_tokens=max_num_batched_tokens,
max_model_len=2048,
gpu_memory_utilization=0.7) as vllm_model:
dynamic_batch_output = vllm_model.generate_greedy(
example_prompts, max_tokens)
with VllmRunner(MODEL,
additional_config={
'SLO_limits_for_dynamic_batch': -1,
},
max_model_len=2048,
gpu_memory_utilization=0.7) as vllm_model:
vllm_output = vllm_model.generate_greedy(example_prompts, max_tokens)
check_outputs_equal(
outputs_0_lst=vllm_output,
outputs_1_lst=dynamic_batch_output,
name_0="vllm_output",
name_1="chunked_prefill_output",
)
def test_async_scheduling_eager() -> None:
prompts = [
"Hello, my name is",
"The president of the United States is",
"The capital of France is",
"The future of AI is",
] * 10
sampling_params = SamplingParams(temperature=0.2,
max_tokens=10,
stop_token_ids=None)
with VllmRunner(
"Qwen/Qwen2.5-0.5B-Instruct",
max_model_len=4096,
max_num_seqs=50,
dtype="bfloat16",
gpu_memory_utilization=0.9,
async_scheduling=True,
) as vllm_model:
vllm_model.generate(prompts, sampling_params=sampling_params)
def test_async_scheduling_with_full_graph() -> None:
prompts = [
"Hello, my name is",
"The president of the United States is",
"The capital of France is",
"The future of AI is",
] * 10
sampling_params = SamplingParams(temperature=0.2,
max_tokens=10,
stop_token_ids=None)
with VllmRunner("Qwen/Qwen3-8B",
max_model_len=4096,
max_num_seqs=50,
dtype="bfloat16",
gpu_memory_utilization=0.9,
async_scheduling=True,
compilation_config={"cudagraph_mode":
"FULL"}) as vllm_model:
vllm_model.generate(prompts, sampling_params=sampling_params)