[CI] Add skipped testcases. (#5254)
### What this PR does / why we need it?
Some E2E testcases are not in our CI workflow, this PR add them back.
- vLLM version: release/v0.13.0
- vLLM main:
ad32e3e19c
Signed-off-by: menogrey <1299267905@qq.com>
This commit is contained in:
@@ -48,9 +48,11 @@ def test_models_with_multistream_overlap_shared_expert(
|
||||
model,
|
||||
max_model_len=1024,
|
||||
enforce_eager=True,
|
||||
cudagraph_capture_sizes=[4, 8, 16, 32],
|
||||
additional_config={
|
||||
"multistream_overlap_shared_expert": True,
|
||||
},
|
||||
quantization="ascend",
|
||||
) as runner:
|
||||
vllm_moe_ms_eager_outputs = runner.model.generate(
|
||||
prompts, sampling_params)
|
||||
@@ -58,9 +60,11 @@ def test_models_with_multistream_overlap_shared_expert(
|
||||
with VllmRunner(
|
||||
model,
|
||||
max_model_len=1024,
|
||||
cudagraph_capture_sizes=[4, 8, 16, 32],
|
||||
additional_config={
|
||||
"multistream_overlap_shared_expert": True,
|
||||
},
|
||||
quantization="ascend",
|
||||
) as runner:
|
||||
vllm_moe_ms_aclgraph_outputs = runner.model.generate(
|
||||
prompts, sampling_params)
|
||||
@@ -69,6 +73,8 @@ def test_models_with_multistream_overlap_shared_expert(
|
||||
model,
|
||||
max_model_len=1024,
|
||||
enforce_eager=True,
|
||||
cudagraph_capture_sizes=[4, 8, 16, 32],
|
||||
quantization="ascend",
|
||||
) as runner:
|
||||
vllm_eager_outputs = runner.model.generate(prompts, sampling_params)
|
||||
|
||||
|
||||
Reference in New Issue
Block a user