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

[LoRA] Change lora_tokenizers capacity (#10796)

Browse Source 
Signed-off-by: Xin Yang <xyang19@gmail.com>
This commit is contained in:
Xin Yang
2024-12-04 09:40:16 -08:00
committed by GitHub
parent c92acb9693
commit 01d079fd8e
7 changed files with 31 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

20
tests/lora/test_tokenizer_group.py
View File

@@ -17,6 +17,7 @@ async def test_tokenizer_group_lora(sql_lora_files, tokenizer_group_type):
tokenizer_id="gpt2",
enable_lora=True,
max_num_seqs=1,
max_loras=1,
max_input_length=None,
)
lora_request = LoRARequest("1", 1, sql_lora_files)
@@ -53,3 +54,22 @@ def test_get_lora_tokenizer(sql_lora_files, tmp_path):
lora_request = LoRARequest("1", 1, str(tmp_path))
tokenizer = get_lora_tokenizer(lora_request)
assert not tokenizer
@pytest.mark.parametrize("enable_lora", [True, False])
@pytest.mark.parametrize("max_num_seqs", [1, 2])
@pytest.mark.parametrize("max_loras", [1, 2])
def test_lora_tokenizers(enable_lora, max_num_seqs, max_loras):
tokenizer_group = get_tokenizer_group(
get_tokenizer_pool_config(None),
tokenizer_id="gpt2",
enable_lora=enable_lora,
max_num_seqs=max_num_seqs,
max_loras=max_loras,
max_input_length=None,
)
if enable_lora:
assert tokenizer_group.lora_tokenizers.capacity == max(
max_num_seqs, max_loras)
else:
assert tokenizer_group.lora_tokenizers.capacity == 0