Deprecate best_of Sampling Parameter in anticipation for vLLM V1 (#13997)

Signed-off-by: vincent-4 <vincentzhongy+githubvincent4@gmail.com>
Signed-off-by: Brayden Zhong <b8zhong@uwaterloo.ca>
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Brayden Zhong <b8zhong@uwaterloo.ca>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>

tests/core/test_scheduler.py

@@ -617,7 +617,6 @@ def test_schedule_decode_blocks_to_copy_update():
                                      num_gpu_blocks=16)
     _, seq_group = create_dummy_prompt("1",
                                        prompt_length=60,
-                                       best_of=2,
                                        block_size=block_size)
     curr_loras = None
     scheduler._allocate_and_set_running(seq_group)
@@ -686,7 +685,6 @@ def test_schedule_swapped_cannot_swap_in():
     for i in range(2):
         _, seq_group = create_dummy_prompt(str(i),
                                            prompt_length=60,
-                                           best_of=2,
                                            block_size=block_size)
         scheduler._allocate_and_set_running(seq_group)
         append_new_token_seq_group(60, seq_group, 1)
@@ -717,7 +715,6 @@ def test_infeasible_swap():
     for i in range(2):
         _, seq_group = create_dummy_prompt(str(i),
                                            prompt_length=60,
-                                           best_of=2,
                                            block_size=block_size)
         scheduler._allocate_and_set_running(seq_group)
         append_new_token_seq_group(60, seq_group, 1)
@@ -747,7 +744,6 @@ def test_schedule_swapped_blocks_to_copy():
     curr_loras = None
     _, seq_group = create_dummy_prompt("1",
                                        prompt_length=60,
-                                       best_of=2,
                                        block_size=block_size)
     scheduler._allocate_and_set_running(seq_group)
     append_new_token_seq_group(60, seq_group, 1)

tests/core/utils.py

@@ -18,7 +18,6 @@ def create_dummy_prompt(
     prompt_length: int = -1,
     block_size: Optional[int] = None,
     lora_request: Optional[LoRARequest] = None,
-    best_of: int = 1,
     prompt_tokens: Optional[list[int]] = None,
     min_tokens: int = 0,
     max_tokens: int = 16,
@@ -32,17 +31,19 @@ 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),
-                      inputs=token_inputs(prompt_tokens, prompt=prompt_str),
-                      block_size=block_size)
-    seq_group = SequenceGroup(request_id=request_id,
-                              seqs=[prompt],
-                              arrival_time=time.time(),
-                              sampling_params=SamplingParams(
-                                  best_of=best_of,
-                                  max_tokens=max_tokens,
-                                  min_tokens=min_tokens),
-                              lora_request=lora_request)
+    prompt = Sequence(
+        int(request_id),
+        inputs=token_inputs(prompt_tokens, prompt=prompt_str),
+        block_size=block_size,
+    )
+    seq_group = SequenceGroup(
+        request_id=request_id,
+        seqs=[prompt],
+        arrival_time=time.time(),
+        sampling_params=SamplingParams(max_tokens=max_tokens,
+                                       min_tokens=min_tokens),
+        lora_request=lora_request,
+    )
 
     return prompt, seq_group
 
@@ -72,7 +73,6 @@ def create_dummy_prompt_encoder_decoder(
     encoder_prompt_length: int,
     block_size: Optional[int] = None,
     lora_request: Optional[LoRARequest] = None,
-    best_of: int = 1,
 ) -> tuple[Sequence, Sequence, SequenceGroup]:
     if not block_size:
         block_size = decoder_prompt_length
@@ -102,7 +102,6 @@ def create_dummy_prompt_encoder_decoder(
 
     seq_group = SequenceGroup(request_id=request_id,
                               seqs=[decoder_prompt],
-                              sampling_params=SamplingParams(best_of=best_of),
                               arrival_time=time.time(),
                               lora_request=lora_request,
                               encoder_seq=encoder_prompt)

tests/v1/sample/test_sampling_params_e2e.py

@@ -25,14 +25,6 @@ def test_n_gt_1(model):
     assert len(outputs[0].outputs) == 3
 
 
-def test_best_of(model):
-    """Raise a ValueError since best_of is deprecated."""
-
-    params = SamplingParams(n=2, best_of=3)
-    with pytest.raises(ValueError):
-        _ = model.generate(PROMPT, params)
-
-
 def test_penalties(model):
     """Check that we do not get errors if applied."""