[Core] Cross-attention KV caching and memory-management (towards even…

…tual encoder/decoder model support) (vllm-project#4837)

tests/core/utils.py

@@ -179,56 +179,5 @@ def create_seq_group_encoder_decoder(
                          encoder_seq=encoder_seq)
 
 
-def create_seq_group_encoder_decoder(
-        seq_prompt_len: int = 1024,
-        seq_output_lens: Iterable[int] = (128, ),
-        request_id: str = '0',
-        seq_id_start: int = 0,
-        sampling_params: Optional[SamplingParams] = None) -> SequenceGroup:
-
-    assert len(seq_output_lens) > 0
-
-    if sampling_params is None:
-        sampling_params = SamplingParams()
-
-    prompt_token_ids = [0] * seq_prompt_len
-
-    seqs = []
-    for seq_id_offset, output_len in enumerate(seq_output_lens):
-        seq = Sequence(
-            seq_id=seq_id_start + seq_id_offset,
-            inputs={
-                "prompt": "",
-                "prompt_token_ids": prompt_token_ids,
-                "multi_modal_data": None,
-            },
-            block_size=16,
-        )
-
-        for i in range(output_len):
-            seq.append_token_id(
-                token_id=i,
-                logprobs={i: Logprob(0.0)},
-            )
-        seqs.append(seq)
-
-    # Encoder sequence
-    encoder_seq = Sequence(
-        seq_id=seq_id_start + len(seq_output_lens),
-        inputs={
-            "prompt": "",
-            "prompt_token_ids": prompt_token_ids,
-            "multi_modal_data": None,
-        },
-        block_size=16,
-    )
-
-    return SequenceGroup(request_id=request_id,
-                         seqs=seqs,
-                         sampling_params=sampling_params,
-                         arrival_time=time.time(),
-                         encoder_seq=encoder_seq)
-
-
 def round_up_to_next_block(seq_len: int, block_size: int) -> int:
     return (seq_len + block_size - 1) // block_size