[Core] Sliding window for block manager v2 (#4545)
Co-authored-by: Ruth Evans <ruthevans@Ruths-MacBook-Pro.local>
This commit is contained in:
@@ -1,3 +1,5 @@
|
||||
from typing import Callable, Iterable, Optional
|
||||
|
||||
import pytest
|
||||
|
||||
from vllm import LLM
|
||||
@@ -40,3 +42,27 @@ def create_llm_generator(common_llm_kwargs, per_test_common_llm_kwargs,
|
||||
for llm in generator_inner():
|
||||
yield llm
|
||||
del llm
|
||||
|
||||
|
||||
def get_text_from_llm_generator(llm_generator: Iterable[LLM],
|
||||
prompts,
|
||||
sampling_params,
|
||||
llm_cb: Optional[Callable[[LLM],
|
||||
None]] = None):
|
||||
for llm in llm_generator:
|
||||
if llm_cb:
|
||||
llm_cb(llm)
|
||||
outputs = llm.generate(prompts, sampling_params, use_tqdm=True)
|
||||
text = [output.outputs[0].text for output in outputs]
|
||||
del llm
|
||||
|
||||
return text
|
||||
|
||||
|
||||
def get_token_ids_from_llm_generator(llm_generator, prompts, sampling_params):
|
||||
for llm in llm_generator:
|
||||
outputs = llm.generate(prompts, sampling_params, use_tqdm=True)
|
||||
token_ids = [output.outputs[0].token_ids for output in outputs]
|
||||
del llm
|
||||
|
||||
return token_ids
|
||||
|
||||
Reference in New Issue
Block a user