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

[CI] refect e2e ci test (#5246)

Browse Source 
### What this PR does / why we need it?
efect e2e ci test：
1. tests/e2e/singlecard/pooling/test_embedding.py: remove the eager
parameter and rename test case
2. tests/e2e/singlecard/pooling/test_scoring.py: Rename test cases
3. tests/e2e/singlecard/pooling/test_classification.py: Rename test case
4. tests/e2e/singlecard/test_quantization.py: remove the eager parameter
and chage model to vllm-ascend/Qwen2.5-0.6B-W8A8 and Rename test case
5. tests/e2e/multicard/test_shared_expert_dp.py: Rename test cases
6. tests/e2e/singlecard/test_sampler.py: Rename test cases
7. tests/e2e/singlecard/test_aclgraph_accuracy.py: Rename test cases
8. tests/e2e/multicard/test_offline_inference_distributed.py: Rename
test cases and remove the eager parameter
9. tests/e2e/multicard/long_sequence/test_accuracy.py: Rename test cases
and remove the eager parameter
10. tests/e2e/multicard/long_sequence/test_basic.py: Rename test cases
and remove the eager parameter
11.tests/e2e/multicard/test_expert_parallel.py:remove the eager
parameter
12.tests/e2e/multicard/test_full_graph_mode.py:remove the eager
parameter
13.tests/e2e/multicard/test_ilama_lora_tp2.py:remove the eager parameter

14.tests/e2e/singlecard/spec_decode_v1/test_v1_mtp_correctness.py:remove
the eager parameter
15.tests/e2e/singlecard/spec_decode_v1/test_v1_spec_decode.py:remove the
eager parameter
16.tests/e2e/singlecard/test_aclgraph_accuracy.py:remove the eager
parameter
17.tests/e2e/singlecard/test_camem.py:remove the eager parameter
18.tests/e2e/singlecard/test_ilama_lora.py:remove the eager parameter

19.tests/e2e/singlecard/test_multistream_overlap_shared_expert.py:remove
the eager parameter
20.tests/e2e/singlecard/test_vlm.py:remove the eager parameter
21.tests/e2e/singlecard/test_xli:remove the eager parameter

### Does this PR introduce _any_ user-facing change?

### How was this patch tested?

- vLLM version: release/v0.13.0
- vLLM main:
ad32e3e19c

Signed-off-by: hfadzxy <starmoon_zhang@163.com>
This commit is contained in:
zhangxinyuehfad
2025-12-23 18:42:35 +08:00
committed by GitHub
parent 5d1f6daef6
commit 8ae7fca947
20 changed files with 61 additions and 88 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
tests/e2e/multicard/long_sequence/test_accuracy.py
View File

@@ -36,7 +36,7 @@ MODELS = [
reason="0.12.0 is not supported for context sequence.")
@pytest.mark.parametrize("model", MODELS)
@pytest.mark.parametrize("max_tokens", [10])
def test_output_between_tp_and_cp(
def test_models_long_sequence_output_between_tp_and_cp(
model: str,
max_tokens: int,
) -> None:
@@ -69,7 +69,6 @@ def test_output_between_tp_and_cp(
"tensor_parallel_size": 1,
"decode_context_parallel_size": 1,
"prefill_context_parallel_size": 2,
"enforce_eager": False,
"compilation_config": {
"cudagraph_mode": "FULL_DECODE_ONLY",
"cudagraph_capture_sizes": [4, 8, 24, 48, 60]

10
tests/e2e/multicard/long_sequence/test_basic.py
View File

@@ -34,7 +34,7 @@ os.environ["HCCL_BUFFSIZE"] = "768"
@pytest.mark.skipif(vllm_version_is('0.12.0'),
reason="0.12.0 is not supported for context sequence.")
def test_pcp_dcp_basic():
def test_models_pcp_dcp_basic():
prompts = [
"The capital of France is", "Hello, my name is Tom, I am",
"The president of United States is", "AI future is"
@@ -69,7 +69,7 @@ def test_pcp_dcp_basic():
@pytest.mark.skipif(vllm_version_is('0.12.0'),
reason="0.12.0 is not supported for context sequence.")
def test_pcp_dcp_full_graph():
def test_models_pcp_dcp_full_graph():
prompts = [
"The capital of France is", "Hello, my name is Tom, I am",
"The president of United States is", "AI future is"
@@ -77,7 +77,6 @@ def test_pcp_dcp_full_graph():
model = "deepseek-ai/DeepSeek-V2-Lite-Chat"
sampling_params = SamplingParams(max_tokens=32, temperature=0.0)
with VllmRunner(model,
enforce_eager=False,
max_model_len=1024,
tensor_parallel_size=2,
prefill_context_parallel_size=2,
@@ -93,7 +92,6 @@ def test_pcp_dcp_full_graph():
model = "vllm-ascend/Qwen3-30B-A3B-W8A8"
with VllmRunner(model,
enforce_eager=False,
max_model_len=1024,
tensor_parallel_size=2,
prefill_context_parallel_size=2,
@@ -110,7 +108,7 @@ def test_pcp_dcp_full_graph():
@pytest.mark.skipif(vllm_version_is('0.12.0'),
reason="0.12.0 is not supported for context sequence.")
def test_pcp_dcp_piece_wise():
def test_models_pcp_dcp_piece_wise():
prompts = [
"The capital of France is", "Hello, my name is Tom, I am",
"The president of United States is", "AI future is"
@@ -118,7 +116,6 @@ def test_pcp_dcp_piece_wise():
model = "deepseek-ai/DeepSeek-V2-Lite-Chat"
sampling_params = SamplingParams(max_tokens=32, temperature=0.0)
with VllmRunner(model,
enforce_eager=False,
max_model_len=1024,
tensor_parallel_size=2,
prefill_context_parallel_size=2,
@@ -130,7 +127,6 @@ def test_pcp_dcp_piece_wise():
model = "vllm-ascend/Qwen3-30B-A3B-W8A8"
with VllmRunner(model,
enforce_eager=False,
max_model_len=1024,
tensor_parallel_size=2,
prefill_context_parallel_size=2,

6
tests/e2e/multicard/test_expert_parallel.py
View File

@@ -15,14 +15,12 @@ def test_deepseek_correctness_ep(model_name):
max_tokens = 5
# FIXME: Really strange that chunked prefill might lead to different results, investigate further
with VllmRunner(model_name, tensor_parallel_size=2,
enforce_eager=False) as vllm_model:
with VllmRunner(model_name, tensor_parallel_size=2) as vllm_model:
tp_output = vllm_model.generate_greedy(example_prompts, max_tokens)
with VllmRunner(model_name,
tensor_parallel_size=2,
enable_expert_parallel=True,
enforce_eager=False) as vllm_model:
enable_expert_parallel=True) as vllm_model:
ep_output = vllm_model.generate_greedy(example_prompts, max_tokens)
check_outputs_equal(

4
tests/e2e/multicard/test_full_graph_mode.py
View File

@@ -41,7 +41,6 @@ def test_qwen3_moe_full_decode_only_tp2():
with VllmRunner(model,
max_model_len=1024,
tensor_parallel_size=2,
enforce_eager=False,
compilation_config={
"cudagraph_mode": "FULL_DECODE_ONLY",
"cudagraph_capture_sizes": [4, 8, 24, 48, 60]
@@ -53,7 +52,6 @@ def test_qwen3_moe_full_decode_only_tp2():
model,
max_model_len=1024,
tensor_parallel_size=2,
enforce_eager=False,
) as runner:
vllm_eager_outputs = runner.model.generate(prompts, sampling_params)
@@ -87,7 +85,6 @@ def test_qwen3_moe_full_graph_tp2():
with VllmRunner(model,
max_model_len=1024,
tensor_parallel_size=2,
enforce_eager=False,
compilation_config={
"cudagraph_mode": "FULL",
"cudagraph_capture_sizes": [4, 8, 24, 48, 60]
@@ -99,7 +96,6 @@ def test_qwen3_moe_full_graph_tp2():
model,
max_model_len=1024,
tensor_parallel_size=2,
enforce_eager=False,
) as runner:
vllm_eager_outputs = runner.model.generate(prompts, sampling_params)

19
tests/e2e/multicard/test_ilama_lora_tp2.py
View File

@@ -8,15 +8,16 @@ from tests.e2e.singlecard.test_ilama_lora import (EXPECTED_LORA_OUTPUT,
@pytest.mark.parametrize("distributed_executor_backend", ["mp"])
def test_ilama_lora_tp2(distributed_executor_backend, ilama_lora_files):
with VllmRunner(snapshot_download(MODEL_PATH),
enable_lora=True,
max_loras=4,
dtype="half",
max_model_len=1024,
max_num_seqs=16,
tensor_parallel_size=2,
distributed_executor_backend=distributed_executor_backend,
enforce_eager=False) as vllm_model:
with VllmRunner(
snapshot_download(MODEL_PATH),
enable_lora=True,
max_loras=4,
dtype="half",
max_model_len=1024,
max_num_seqs=16,
tensor_parallel_size=2,
distributed_executor_backend=distributed_executor_backend,
) as vllm_model:
output = do_sample(vllm_model.model, ilama_lora_files, lora_id=2)
for i in range(len(EXPECTED_LORA_OUTPUT)):

2
tests/e2e/multicard/test_offline_inference_distributed.py
View File

@@ -189,7 +189,6 @@ def test_qwen3_dense_fc1_tp2(model):
with VllmRunner(
snapshot_download(model),
max_model_len=8192,
enforce_eager=False,
dtype="auto",
tensor_parallel_size=2,
quantization="ascend",
@@ -209,7 +208,6 @@ def test_qwen3_dense_prefetch_mlp_weight_tp2(model):
with VllmRunner(
snapshot_download(model),
max_model_len=8192,
enforce_eager=False,
dtype="auto",
tensor_parallel_size=2,
quantization="ascend",