mirror of
https://git.datalinker.icu/vllm-project/vllm.git
synced 2025-12-11 07:25:01 +08:00
[CI/Build] Fix LoRA OOM (#16624)
Signed-off-by: Jee Jee Li <pandaleefree@gmail.com>
This commit is contained in:
parent
6ae996a873
commit
1575c1701a
@ -66,8 +66,12 @@ def test_minicpmv_lora(minicpmv_lora_files):
|
||||
max_loras=2,
|
||||
max_lora_rank=8,
|
||||
enforce_eager=True,
|
||||
max_model_len=2048,
|
||||
limit_mm_per_prompt={
|
||||
"image": 2,
|
||||
"video": 0
|
||||
},
|
||||
trust_remote_code=True,
|
||||
enable_chunked_prefill=True,
|
||||
)
|
||||
output1 = do_sample(llm, minicpmv_lora_files, lora_id=1)
|
||||
for i in range(len(EXPECTED_OUTPUT)):
|
||||
@ -91,9 +95,11 @@ def test_minicpmv_tp4_wo_fully_sharded_loras(minicpmv_lora_files):
|
||||
max_loras=4,
|
||||
max_lora_rank=64,
|
||||
tensor_parallel_size=4,
|
||||
limit_mm_per_prompt={
|
||||
"image": 2,
|
||||
"video": 0
|
||||
},
|
||||
trust_remote_code=True,
|
||||
enforce_eager=True,
|
||||
enable_chunked_prefill=True,
|
||||
)
|
||||
output_tp = do_sample(llm, minicpmv_lora_files, lora_id=1)
|
||||
for i in range(len(EXPECTED_OUTPUT)):
|
||||
@ -115,8 +121,11 @@ def test_minicpmv_tp4_fully_sharded_loras(minicpmv_lora_files):
|
||||
max_lora_rank=8,
|
||||
tensor_parallel_size=4,
|
||||
trust_remote_code=True,
|
||||
limit_mm_per_prompt={
|
||||
"image": 1,
|
||||
"video": 0
|
||||
},
|
||||
fully_sharded_loras=True,
|
||||
enable_chunked_prefill=True,
|
||||
)
|
||||
output_tp = do_sample(llm, minicpmv_lora_files, lora_id=1)
|
||||
for i in range(len(EXPECTED_OUTPUT)):
|
||||
|
||||
Loading…
x
Reference in New Issue
Block a user