A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 16, 15:45 UTC
Fix test_schedule_swapped_simple in test_scheduler.py (#8780)
sroy745 committed
Sep 24, 2024 at 21:26 UTC
ee777d9c30418ffa9d98f98dd27c0ddea346c49c
1 file changed
+10
-4
tests/core/test_scheduler.py
+10
-4
@@ -747,13 +747,19 @@ def test_schedule_decode_blocks_to_copy_update(use_v2_block_manager: bool):
747
assert output.blocks_to_copy == [(2, 3)]
748
749
750
-def test_schedule_swapped_simple():
751
- scheduler = initialize_scheduler()
750
+@pytest.mark.parametrize('use_v2_block_manager', [True, False])
751
+def test_schedule_swapped_simple(use_v2_block_manager: bool):
752
+ block_size = 4
753
+ scheduler = initialize_scheduler(use_v2_block_manager=use_v2_block_manager,
754
+ block_size=block_size)
755
curr_loras = None
756
blocks_to_swap_out: List[Tuple[int, int]] = []
754
- _, seq_group = create_dummy_prompt("1", prompt_length=60, best_of=2)
757
+ _, seq_group = create_dummy_prompt("1",
758
+ prompt_length=4,
759
+ best_of=2,
760
+ block_size=block_size)
761
scheduler._allocate_and_set_running(seq_group)
756
- append_new_token_seq_group(60, seq_group, 1)
762
+ append_new_token_seq_group(4, seq_group, 1)
763
scheduler._swap_out(seq_group, blocks_to_swap_out)
764
scheduler._add_seq_group_to_swapped(seq_group)
765