[Refactor] Remove dead or duplicate tests (#58446)
Signed-off-by: yewentao256 <zhyanwentao@126.com>
This commit is contained in:
@@ -171,42 +171,6 @@ def test_connector_metadata_precedes_async_placeholder_advance(monkeypatch):
|
||||
assert observed_placeholders == [0, 1]
|
||||
|
||||
|
||||
def test_preempt():
|
||||
scheduler = create_scheduler(async_scheduling=True)
|
||||
requests = create_requests(num_requests=10, max_tokens=20)
|
||||
|
||||
for req in requests:
|
||||
scheduler.add_request(req)
|
||||
|
||||
sched_outputs: deque[SchedulerOutput] = deque()
|
||||
sched_outputs.append(scheduler.schedule())
|
||||
sched_outputs.append(scheduler.schedule())
|
||||
|
||||
abort_order = [0, 8, 3, 1, 6, 4, 2, 5, 7, 9]
|
||||
abort_order_copy = abort_order.copy()
|
||||
|
||||
def abort_request():
|
||||
if not abort_order:
|
||||
return
|
||||
req = requests[abort_order.pop(0)]
|
||||
scheduler.finish_requests(req.request_id, RequestStatus.FINISHED_ABORTED)
|
||||
|
||||
while sched_outputs:
|
||||
# Abort a scheduled request.
|
||||
abort_request()
|
||||
sched_output = sched_outputs.popleft()
|
||||
model_runner_output = _make_model_runner_output(sched_output)
|
||||
scheduler.update_from_output(sched_output, model_runner_output)
|
||||
|
||||
sched_output = scheduler.schedule()
|
||||
if sched_output.num_scheduled_tokens:
|
||||
sched_outputs.append(sched_output)
|
||||
|
||||
for i, req in enumerate(requests):
|
||||
assert req.status == RequestStatus.FINISHED_ABORTED
|
||||
assert req.num_output_tokens == abort_order_copy.index(i)
|
||||
|
||||
|
||||
def test_prefix_caching_for_prefill_dedup():
|
||||
CHUNK_SIZE = 1000
|
||||
BLOCK_SIZE = 16
|
||||
|
||||
@@ -16,38 +16,6 @@ def new_kv_cache_spec():
|
||||
)
|
||||
|
||||
|
||||
def test_initialize_kv_cache_for_kv_sharing_different_attn_groups():
|
||||
"""Test initializing KV cache sharing with different attention groups.
|
||||
Layers in the same KV cache group might be placed in different attn groups
|
||||
if they have different attention backends.
|
||||
"""
|
||||
shared_kv_cache_layers = {
|
||||
"model.layers.2": "model.layers.0",
|
||||
"model.layers.3": "model.layers.1",
|
||||
}
|
||||
|
||||
# Layers 0 and 1 both belong in KV cache group 0
|
||||
# However, if they have different attention backends, they will be
|
||||
# placed in different attention groups for KV cache group 0
|
||||
kv_cache_groups = [
|
||||
KVCacheGroupSpec(["model.layers.0", "model.layers.1"], new_kv_cache_spec()),
|
||||
]
|
||||
|
||||
add_kv_sharing_layers_to_kv_cache_groups(
|
||||
shared_kv_cache_layers=shared_kv_cache_layers,
|
||||
kv_cache_groups=kv_cache_groups,
|
||||
)
|
||||
|
||||
# Check that the layers were added to the correct KV cache group
|
||||
assert len(kv_cache_groups) == 1
|
||||
assert kv_cache_groups[0].layer_names == [
|
||||
"model.layers.0",
|
||||
"model.layers.1",
|
||||
"model.layers.2",
|
||||
"model.layers.3",
|
||||
]
|
||||
|
||||
|
||||
def test_initialize_kv_cache_for_kv_sharing_same_attn_groups():
|
||||
"""Test case assuming that all layers in the same KV cache group have the same
|
||||
attention backends. This is true for most models.
|
||||
|
||||
Reference in New Issue
Block a user