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

[LoRA] Cleanup LoRA unused code (#29611)

Browse Source 
Signed-off-by: Jee Jee Li <pandaleefree@gmail.com>
Co-authored-by: Cyrus Leung <tlleungac@connect.ust.hk>
This commit is contained in:
Jee Jee Li
2025-11-29 14:52:58 +08:00
committed by GitHub
parent 4a80ad0a25
commit 39e63dec7c
46 changed files with 126 additions and 173 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
tests/lora/test_olmoe_tp.py
View File

@@ -40,7 +40,10 @@ EXPECTED_BASE_MODEL_OUTPUT = [
def generate_and_test(
llm: vllm.LLM, lora_path: str, lora_id: list[int | None] | int | None
llm: vllm.LLM,
lora_path: str,
lora_id: list[int | None] | int | None,
compare_lower: bool = False,
) -> None:
prompts = [
PROMPT_TEMPLATE.format(context="How many candidates are there?"),
@@ -74,12 +77,18 @@ def generate_and_test(
for i in range(len(EXPECTED_LORA_OUTPUT)):
req_lora_id = lora_id[i] if isinstance(lora_id, list) else lora_id
generated_text = generated_texts[i]
expected_output = (
EXPECTED_LORA_OUTPUT[i]
if req_lora_id is not None
else EXPECTED_BASE_MODEL_OUTPUT[i]
)
assert generated_texts[i].startswith(expected_output)
if compare_lower:
generated_text = generated_text.lower()
expected_output = expected_output.lower()
assert generated_text.startswith(expected_output)
def test_olmoe_lora(olmoe_lora_files):
@@ -146,6 +155,9 @@ def test_olmoe_lora_tp4(olmoe_lora_files, fully_sharded_loras):
tensor_parallel_size=4,
fully_sharded_loras=fully_sharded_loras,
)
generate_and_test(llm, olmoe_lora_files, lora_id=1)
generate_and_test(llm, olmoe_lora_files, lora_id=2)
generate_and_test(
llm, olmoe_lora_files, lora_id=1, compare_lower=fully_sharded_loras
)
generate_and_test(
llm, olmoe_lora_files, lora_id=2, compare_lower=fully_sharded_loras
)