Skip to content

[CI] Add long and short prompt tests for DeepSeek-V3.2 #6499

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
wangxiyuan merged 1 commit into from
Feb 4, 2026
Merged

[CI] Add long and short prompt tests for DeepSeek-V3.2 #6499

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
14 changes: 10 additions & 4 deletions tests/e2e/multicard/2-cards/test_offline_inference_distributed.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -245,14 +245,19 @@ def test_qwen3_dense_prefetch_mlp_weight_tp2(model):
@patch.dict(os.environ, {"ASCEND_AGGREGATE_ENABLE": "1"})
@patch.dict(os.environ, {"HCCL_BUFFSIZE": "1024"})
def test_deepseek3_2_w8a8_pruning_mtp_tp2_ep():
example_prompts = [
"Hello, my name is",
short_example_prompts = [
"Hello ",
]
max_tokens = 5
# "max_position_embeddings": 163840,
long_example_prompts = [
"Hello " * (163839 - 500) + "Hello"
]
Comment on lines +252 to +254
Copy link
Copy Markdown
Contributor

@gemini-code-assist gemini-code-assist Bot Feb 3, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

critical

The long_example_prompts list is empty. As a result, the call to vllm_model.generate_greedy(long_example_prompts, max_tokens) on line 274 will be a no-op. To make the test meaningful, this list should be populated with long prompts.

max_tokens = 500
with VllmRunner("vllm-ascend/DeepSeek-V3.2-W8A8-Pruning",
tensor_parallel_size=2,
quantization="ascend",
enable_expert_parallel=True,
max_model_len=163840,
compilation_config={
"cudagraph_capture_sizes": [3, 6, 9, 12],
"cudagraph_mode": "FULL_DECODE_ONLY"
Expand All @@ -266,7 +271,8 @@ def test_deepseek3_2_w8a8_pruning_mtp_tp2_ep():
},
reasoning_parser="deepseek_v3",
tokenizer_mode="deepseek_v32") as vllm_model:
vllm_model.generate_greedy(example_prompts, max_tokens)
vllm_model.generate_greedy(short_example_prompts, max_tokens)
vllm_model.generate_greedy(long_example_prompts, max_tokens)


@pytest.mark.parametrize("model", QWEN_W4A4_MODELS)
Expand Down
Loading