biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[CI/Test] improve robustness of test (vllm_runner) (#5357)

Browse Source 
[CI/Test] improve robustness of test by replacing del with context manager (vllm_runner) (#5357)
This commit is contained in:
youkaichao
2024-06-08 01:59:20 -07:00
committed by GitHub
parent 9fb900f90c
commit 8ea5e44a43
28 changed files with 431 additions and 470 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
tests/models/test_models.py
View File

@@ -37,9 +37,8 @@ def test_models(
with hf_runner(model, dtype=dtype) as hf_model:
hf_outputs = hf_model.generate_greedy(example_prompts, max_tokens)
vllm_model = vllm_runner(model, dtype=dtype)
vllm_outputs = vllm_model.generate_greedy(example_prompts, max_tokens)
del vllm_model
with vllm_runner(model, dtype=dtype) as vllm_model:
vllm_outputs = vllm_model.generate_greedy(example_prompts, max_tokens)
for i in range(len(example_prompts)):
hf_output_ids, hf_output_str = hf_outputs[i]
@@ -57,9 +56,8 @@ def test_model_print(
model: str,
dtype: str,
) -> None:
vllm_model = vllm_runner(model, dtype=dtype)
# This test is for verifying whether the model's extra_repr
# can be printed correctly.
print(vllm_model.model.llm_engine.model_executor.driver_worker.
model_runner.model)
del vllm_model
with vllm_runner(model, dtype=dtype) as vllm_model:
# This test is for verifying whether the model's extra_repr
# can be printed correctly.
print(vllm_model.model.llm_engine.model_executor.driver_worker.
model_runner.model)