[Model][0/N] Improve all pooling task | clean up (#25817)
Signed-off-by: wang.yuqi <noooop@126.com>
This commit is contained in:
@@ -58,7 +58,9 @@ def test_pooling_params(llm: LLM):
|
||||
)
|
||||
|
||||
|
||||
@pytest.mark.skip_global_cleanup
|
||||
def test_encode_api(llm: LLM):
|
||||
# chunked prefill does not support all pooling
|
||||
err_msg = "pooling_task must be one of.+"
|
||||
with pytest.raises(ValueError, match=err_msg):
|
||||
llm.encode(prompts, use_tqdm=False)
|
||||
|
||||
@@ -35,7 +35,6 @@ def llm():
|
||||
cleanup_dist_env_and_memory()
|
||||
|
||||
|
||||
@pytest.mark.skip_global_cleanup
|
||||
def test_pooling_params(llm: LLM):
|
||||
def get_outputs(normalize):
|
||||
outputs = llm.embed(
|
||||
|
||||
@@ -74,7 +74,6 @@ def test_multiple_pooling_params(llm: LLM):
|
||||
assert len(PROMPTS) == len(outputs)
|
||||
|
||||
|
||||
@pytest.mark.skip_global_cleanup
|
||||
def test_right_side_truncation(llm: LLM):
|
||||
# Embeddings models should truncate the end of the prompt
|
||||
tokenizer = llm.get_tokenizer()
|
||||
|
||||
@@ -33,7 +33,6 @@ def llm():
|
||||
cleanup_dist_env_and_memory()
|
||||
|
||||
|
||||
@pytest.mark.skip_global_cleanup
|
||||
def test_pooling_params(llm: LLM):
|
||||
def get_outputs(activation):
|
||||
text_1 = "What is the capital of France?"
|
||||
|
||||
Reference in New Issue
Block a user