[CI] Nits for bad initialization of SeqGroup in testing (#4748)

tests/core/test_block_manager.py

@@ -142,8 +142,10 @@ def test_append_slot_cow():
     child = prompt.fork(new_seq_id=2)
 
     # Allocate space for the sequence group.
-    seq_group = SequenceGroup("1", [prompt, child], SamplingParams(),
-                              time.time(), time.perf_counter)
+    seq_group = SequenceGroup(request_id="1",
+                              seqs=[prompt, child],
+                              arrival_time=time.time(),
+                              sampling_params=SamplingParams())
     block_manager.allocate(seq_group)
 
     # Fork and append a new token id. We expect a COW to be scheduled.
@@ -303,8 +305,11 @@ def test_sliding_window_multi_seq():
     assert block_manager.get_num_free_gpu_blocks() == num_gpu_blocks
 
     parent = Sequence(1, "one two three", [0, 1, 2], block_size)
-    seq_group = SequenceGroup("1", [parent], SamplingParams(), time.time(),
-                              None)
+    seq_group = SequenceGroup(request_id="1",
+                              seqs=[parent],
+                              arrival_time=time.time(),
+                              sampling_params=SamplingParams(),
+                              lora_request=None)
     block_manager.allocate(seq_group)
 
     # assert the number of blocks allocated is correct

tests/core/utils.py

@@ -22,10 +22,13 @@ def create_dummy_prompt(
     prompt_tokens = list(range(prompt_length))
     prompt_str = " ".join([str(t) for t in prompt_tokens])
     prompt = Sequence(int(request_id), prompt_str, prompt_tokens, block_size)
-    seq_group = SequenceGroup(
-        request_id, [prompt],
-        SamplingParams(use_beam_search=use_beam_search, best_of=best_of),
-        time.time(), lora_request)
+    seq_group = SequenceGroup(request_id=request_id,
+                              seqs=[prompt],
+                              arrival_time=time.time(),
+                              sampling_params=SamplingParams(
+                                  use_beam_search=use_beam_search,
+                                  best_of=best_of),
+                              lora_request=lora_request)
 
     return prompt, seq_group