[Feature] [Spec decode]: Combine chunked prefill with speculative decoding (#9291)

Signed-off-by: NickLucche <nlucches@redhat.com>

tests/spec_decode/e2e/test_compatibility.py

@@ -5,40 +5,6 @@ from vllm import SamplingParams
 from .conftest import get_output_from_llm_generator
 
 
-@pytest.mark.parametrize("common_llm_kwargs", [{
-    "model": "JackFram/llama-68m",
-    "speculative_model": "JackFram/llama-68m",
-    "num_speculative_tokens": 5,
-}])
-@pytest.mark.parametrize("per_test_common_llm_kwargs", [
-    {
-        "enable_chunked_prefill": True,
-    },
-])
-@pytest.mark.parametrize("test_llm_kwargs", [{}])
-@pytest.mark.parametrize("seed", [1])
-def test_spec_decode_xfail_chunked_prefill(test_llm_generator):
-    """Verify that speculative decoding with chunked prefill fails.
-    """
-    output_len = 128
-    temperature = 0.0
-
-    prompts = [
-        "Hello, my name is",
-    ]
-
-    sampling_params = SamplingParams(
-        max_tokens=output_len,
-        ignore_eos=True,
-        temperature=temperature,
-    )
-
-    with pytest.raises(ValueError,
-                       match="Speculative decoding and chunked prefill"):
-        get_output_from_llm_generator(test_llm_generator, prompts,
-                                      sampling_params)
-
-
 @pytest.mark.parametrize("common_llm_kwargs", [{
     "model": "meta-llama/Llama-2-7b-chat-hf",
     "speculative_model": "JackFram/llama-68m",

tests/spec_decode/e2e/test_multistep_correctness.py

@@ -62,6 +62,16 @@ from .conftest import (get_output_from_llm_generator,
         {
             "speculative_model": "JackFram/llama-68m",
             "num_speculative_tokens": 5,
+            "enable_chunked_prefill": False,
+        },
+        {
+            # Chunked prefill enabled with small value
+            # to make sure we get mixed batches.
+            "speculative_model": "JackFram/llama-68m",
+            "num_speculative_tokens": 5,
+            "enable_chunked_prefill": True,
+            "max_num_batched_tokens": 4,
+            "max_num_seqs": 4
         },
         {
             # Verify the detokenizer assertions in the test work when spec
@@ -141,6 +151,14 @@ def test_spec_decode_e2e_with_detokenization(test_llm_generator,
     {
         "speculative_model": "JackFram/llama-68m",
         "num_speculative_tokens": 5,
+        "enable_chunked_prefill": False,
+    },
+    {
+        "speculative_model": "JackFram/llama-68m",
+        "num_speculative_tokens": 5,
+        "enable_chunked_prefill": True,
+        "max_num_batched_tokens": 4,
+        "max_num_seqs": 4,
     },
 ])
 @pytest.mark.parametrize(
@@ -204,6 +222,14 @@ def test_spec_decode_e2e_greedy_correctness_tiny_model_bs1(
     {
         "speculative_model": "JackFram/llama-68m",
         "num_speculative_tokens": 5,
+        "enable_chunked_prefill": False,
+    },
+    {
+        "speculative_model": "JackFram/llama-68m",
+        "num_speculative_tokens": 5,
+        "enable_chunked_prefill": True,
+        "max_num_batched_tokens": 4,
+        "max_num_seqs": 4
     },
 ])
 @pytest.mark.parametrize(
@@ -255,6 +281,14 @@ def test_spec_decode_e2e_greedy_correctness_tiny_model_large_bs(
     {
         "speculative_model": "JackFram/llama-68m",
         "num_speculative_tokens": 5,
+        "enable_chunked_prefill": False,
+    },
+    {
+        "speculative_model": "JackFram/llama-68m",
+        "num_speculative_tokens": 5,
+        "enable_chunked_prefill": True,
+        "max_num_batched_tokens": 4,
+        "max_num_seqs": 4
     },
 ])
 @pytest.mark.parametrize("max_output_len", [
@@ -300,6 +334,14 @@ def test_spec_decode_e2e_greedy_correctness_tiny_model_large_bs_diff_output_len(
     {
         "speculative_model": "JackFram/llama-68m",
         "num_speculative_tokens": 5,
+        "enable_chunked_prefill": False,
+    },
+    {
+        "speculative_model": "JackFram/llama-68m",
+        "num_speculative_tokens": 5,
+        "enable_chunked_prefill": True,
+        "max_num_batched_tokens": 4,
+        "max_num_seqs": 4
     },
 ])
 @pytest.mark.parametrize("batch_size", [1])
@@ -347,6 +389,14 @@ def test_spec_decode_e2e_greedy_correctness_real_model_bs1(
     {
         "speculative_model": "JackFram/llama-68m",
         "num_speculative_tokens": 5,
+        "enable_chunked_prefill": False,
+    },
+    {
+        "speculative_model": "JackFram/llama-68m",
+        "num_speculative_tokens": 5,
+        "enable_chunked_prefill": True,
+        "max_num_batched_tokens": 4,
+        "max_num_seqs": 4
     },
 ])
 @pytest.mark.parametrize("batch_size", [32])
@@ -397,6 +447,14 @@ def test_spec_decode_e2e_greedy_correctness_real_model_large_bs(
     {
         "speculative_model": "JackFram/llama-68m",
         "num_speculative_tokens": 5,
+        "enable_chunked_prefill": False,
+    },
+    {
+        "speculative_model": "JackFram/llama-68m",
+        "num_speculative_tokens": 5,
+        "enable_chunked_prefill": True,
+        "max_num_batched_tokens": 4,
+        "max_num_seqs": 4
     },
 ])
 @pytest.mark.parametrize(
@@ -454,6 +512,14 @@ def test_spec_decode_e2e_greedy_correctness_with_preemption(
     {
         "speculative_model": "JackFram/llama-68m",
         "num_speculative_tokens": 5,
+        "enable_chunked_prefill": False,
+    },
+    {
+        "speculative_model": "JackFram/llama-68m",
+        "num_speculative_tokens": 5,
+        "enable_chunked_prefill": True,
+        "max_num_batched_tokens": 4,
+        "max_num_seqs": 4
     },
 ])
 @pytest.mark.parametrize("batch_size", [2])
@@ -503,6 +569,15 @@ def test_spec_decode_different_block_size(vllm_runner, common_llm_kwargs,
             # Artificially limit the draft model max model len; this forces vLLM
             # to skip speculation once the sequences grow beyond 32-k tokens.
             "speculative_max_model_len": 32,
+            "enable_chunked_prefill": False,
+        },
+        {
+            "speculative_model": "JackFram/llama-68m",
+            "num_speculative_tokens": 5,
+            "enable_chunked_prefill": True,
+            "max_num_batched_tokens": 4,
+            "max_num_seqs": 4,
+            "speculative_max_model_len": 32,
         },
     ])
 @pytest.mark.parametrize("batch_size", [8])
@@ -551,6 +626,15 @@ def test_skip_speculation(vllm_runner, common_llm_kwargs,
         "speculative_model": "JackFram/llama-68m",
         "num_speculative_tokens": 5,
         "speculative_disable_by_batch_size": 2,
+        "enable_chunked_prefill": False,
+    },
+    {
+        "speculative_model": "JackFram/llama-68m",
+        "num_speculative_tokens": 5,
+        "speculative_disable_by_batch_size": 2,
+        "enable_chunked_prefill": True,
+        "max_num_batched_tokens": 4,
+        "max_num_seqs": 4,
     },
 ])
 @pytest.mark.parametrize("batch_size", [8])
@@ -590,10 +674,17 @@ def test_disable_speculation(vllm_runner, common_llm_kwargs,
         {
             "speculative_model": "JackFram/llama-68m",
             "num_speculative_tokens": k,
+            "enable_chunked_prefill": False,
         }
         # Try a range of common k, as well as large speculation.
         for k in [1, 2, 3, 4, 5, 6, 7, 8, 9, 63]
-    ])
+    ] + [{
+        "speculative_model": "JackFram/llama-68m",
+        "num_speculative_tokens": k,
+        "enable_chunked_prefill": True,
+        "max_num_batched_tokens": 4,
+        "max_num_seqs": 4,
+    } for k in [1, 2, 3, 4, 5, 6, 7, 8, 9, 63]])
 @pytest.mark.parametrize("batch_size", [2])
 @pytest.mark.parametrize(
     "output_len",
@@ -636,11 +727,19 @@ def test_many_k(vllm_runner, common_llm_kwargs, per_test_common_llm_kwargs,
         {
             "speculative_model": "JackFram/llama-68m",
             "num_speculative_tokens": k,
-            "spec_decoding_acceptance_method": "typical_acceptance_sampler"
+            "spec_decoding_acceptance_method": "typical_acceptance_sampler",
+            "enable_chunked_prefill": False
         }
         # Try a range of common k.
         for k in [1, 2, 3]
-    ])
+    ] + [{
+        "speculative_model": "JackFram/llama-68m",
+        "num_speculative_tokens": k,
+        "spec_decoding_acceptance_method": "typical_acceptance_sampler",
+        "enable_chunked_prefill": True,
+        "max_num_batched_tokens": 4,
+        "max_num_seqs": 4
+    } for k in [1, 2, 3]])
 @pytest.mark.parametrize("batch_size", [1, 32])
 @pytest.mark.parametrize(
     "output_len",

tests/spec_decode/e2e/test_ngram_correctness.py

@@ -50,18 +50,33 @@ from .conftest import run_equality_correctness_test
         "num_speculative_tokens": 5,
         "ngram_prompt_lookup_max": 3,
     },
+    {
+        "speculative_model": "[ngram]",
+        "num_speculative_tokens": 5,
+        "ngram_prompt_lookup_max": 3,
+    },
 ])
 @pytest.mark.parametrize("output_len", [
     256,
 ])
 @pytest.mark.parametrize("batch_size", [1, 32])
+@pytest.mark.parametrize("prefill_chunk_size", [-1, 4])
 @pytest.mark.parametrize("seed", [1])
 def test_ngram_e2e_greedy_correctness(vllm_runner, common_llm_kwargs,
                                       per_test_common_llm_kwargs,
                                       baseline_llm_kwargs, test_llm_kwargs,
                                       batch_size: int, output_len: int,
-                                      seed: int):
+                                      prefill_chunk_size: int, seed: int):
     """Verify greedy equality on a tiny model with different batch size."""
+    if prefill_chunk_size > 0:
+        common_llm_kwargs.update(
+            **{
+                "enable_chunked_prefill": True,
+                "max_num_batched_tokens": prefill_chunk_size,
+                "max_num_seqs": prefill_chunk_size
+            })
+    else:
+        common_llm_kwargs["enable_chunked_prefill"] = False
     run_equality_correctness_test(vllm_runner,
                                   common_llm_kwargs,
                                   per_test_common_llm_kwargs,
@@ -151,6 +166,16 @@ def test_ngram_e2e_greedy_logprobs(vllm_runner, common_llm_kwargs,
         "speculative_model": "[ngram]",
         "num_speculative_tokens": 5,
         "ngram_prompt_lookup_max": 3,
+        "enable_chunked_prefill": False,
+    },
+    {
+        "speculative_model": "[ngram]",
+        "num_speculative_tokens": 5,
+        "ngram_prompt_lookup_max": 3,
+        "enable_chunked_prefill": True,
+        "speculative_disable_mqa_scorer": True,
+        "max_num_batched_tokens": 4,
+        "max_num_seqs": 4
     },
 ])
 @pytest.mark.parametrize(
@@ -251,6 +276,15 @@ def test_ngram_different_k(vllm_runner, common_llm_kwargs,
                              "num_speculative_tokens": 5,
                              "ngram_prompt_lookup_max": 3,
                              "speculative_disable_by_batch_size": 4
+                         }, {
+                             "speculative_model": "[ngram]",
+                             "num_speculative_tokens": 5,
+                             "ngram_prompt_lookup_max": 3,
+                             "speculative_disable_by_batch_size": 4,
+                             "enable_chunked_prefill": True,
+                             "speculative_disable_mqa_scorer": True,
+                             "max_num_batched_tokens": 4,
+                             "max_num_seqs": 4
                          }])
 @pytest.mark.parametrize("batch_size", [1, 5])
 @pytest.mark.parametrize(

tests/spec_decode/test_ngram_worker.py

@@ -118,7 +118,8 @@ def test_ngram_algo_correctness_for_batches_not_match_all():
         num_gpu_blocks,
         block_size,
         final_prompt_lens=final_prompt_lens)
-
+    for sg in seq_group_metadata_list:
+        sg.is_prompt = False
     proposals = proposer.get_spec_proposals(
         execute_model_req=ExecuteModelRequest(
             seq_group_metadata_list=seq_group_metadata_list,
@@ -147,7 +148,7 @@ def test_ngram_algo_correctness_for_batches_not_match_all():
 def test_ngram_algo_correctness_for_batches_match_all():
     """Verify our ngram algo find the right candidate in the prompt
 
-    For the scenario find candidate in all batchs
+    For the scenario find candidate in all batches
     """
 
     block_size = 32
@@ -192,6 +193,10 @@ def test_ngram_algo_correctness_for_batches_match_all():
         block_size,
         final_prompt_lens=final_prompt_lens)
 
+    # Normally drafter is run on decode requests only; here we check the output
+    # of the ngram worker as it is the sole proposer that has no forward.
+    for sg in seq_group_metadata_list:
+        sg.is_prompt = False
     proposals = proposer.get_spec_proposals(
         execute_model_req=ExecuteModelRequest(
             seq_group_metadata_list=seq_group_metadata_list,

tests/spec_decode/test_scorer.py

@@ -46,12 +46,14 @@ def assert_score_equal(score1: SpeculativeScores,
 @pytest.mark.parametrize('max_propose_len', [1, 3, 5])
 @pytest.mark.parametrize('mixed_propose_len', [True])
 @pytest.mark.parametrize('device', ['cuda'])
+@pytest.mark.parametrize('prefill_chunking', [False, True])
 def test_scorer(model_name: str, batch_size: int, max_propose_len: int,
-                mixed_propose_len: bool, device: str) -> None:
+                mixed_propose_len: bool, device: str,
+                prefill_chunking: bool) -> None:
     """
     Compare the batch expansion scorer and mqa scorer return the same score.
     We test for both queries with the same propose length and different 
-    propose length.
+    propose length, as well as mixed prefill-decode batches.
     """
     seed = 0
     block_size = 32
@@ -67,16 +69,37 @@ def test_scorer(model_name: str, batch_size: int, max_propose_len: int,
     if not mixed_propose_len:
         propose_lens = [max_propose_len] * batch_size
     else:
-        non_zero_cnt = random.randint(0, batch_size)
+        # There must be at least 1 decode request, otherwise
+        # we have nothing to score (`_run_no_spec`).
+        non_zero_cnt = random.randint(1, batch_size)
         propose_lens = [max_propose_len
                         ] * non_zero_cnt + [0] * (batch_size - non_zero_cnt)
         random.shuffle(propose_lens)
 
-    proposals = create_proposal(propose_lens, vocab_size, device)
     seq_group_metadatalist, _, _ = create_batch(batch_size,
                                                 max_propose_len,
                                                 block_size=block_size,
                                                 num_gpu_blocks=num_gpu_blocks)
+
+    if mixed_propose_len and prefill_chunking and (n_prefills :=
+                                                   batch_size - non_zero_cnt):
+        prefill, _, _ = create_batch(n_prefills,
+                                     None,
+                                     prefill_chunk_size=4,
+                                     block_size=block_size,
+                                     num_gpu_blocks=num_gpu_blocks,
+                                     seq_ids=list(
+                                         range(batch_size,
+                                               batch_size + n_prefills)))
+        # re-order to guarantee prefill|decode order
+        target_group_metadatalist = [
+            seq_group_metadatalist[i] for i, p in enumerate(propose_lens)
+            if p > 0
+        ]
+        seq_group_metadatalist = prefill + target_group_metadatalist
+        propose_lens = [0] * n_prefills + [p for p in propose_lens if p > 0]
+
+    proposals = create_proposal(propose_lens, vocab_size, device)
     requests = ExecuteModelRequest(seq_group_metadatalist,
                                    num_lookahead_slots=max_propose_len)
 

tests/spec_decode/test_spec_decode_worker.py

@@ -10,6 +10,7 @@ import torch
 from vllm.model_executor.layers.sampler import SamplerOutput
 from vllm.model_executor.utils import set_random_seed
 from vllm.sequence import ExecuteModelRequest, SequenceOutput
+from vllm.spec_decode.batch_expansion import BatchExpansionTop1Scorer
 from vllm.spec_decode.interfaces import SpeculativeProposals
 from vllm.spec_decode.metrics import (AsyncMetricsCollector,
                                       SpecDecodeWorkerMetrics)
@@ -819,3 +820,84 @@ def test_handle_finished_requests():
     # and 'request-3' are removed from seq_with_bonus_token_in_last_step.
     assert worker._seq_with_bonus_token_in_last_step == \
         {4,5,10}
+
+
+@pytest.mark.parametrize('k', [3])
+@pytest.mark.parametrize('batch_size', [2, 32])
+@pytest.mark.parametrize("batch_composition",
+                         ["prefill_only", "decode_only", "mixed"])
+@torch.inference_mode()
+def test_chunked_prefill_flow(k: int, batch_size: int, batch_composition: str):
+    """
+        Verify SpecDecodeWorker calls match the expected flow.
+    """
+    vocab_size = 32_000
+    draft_worker = mock_worker(cls=MultiStepWorker)
+    target_worker = mock_worker()
+    metrics_collector = MagicMock(spec=AsyncMetricsCollector)
+    worker = SpecDecodeWorker(draft_worker,
+                              target_worker,
+                              mock_spec_decode_sampler("rejection_sampler"),
+                              disable_logprobs=False,
+                              metrics_collector=metrics_collector)
+    exception_secret = 'artificial stop'
+    worker.scorer = mock_worker(BatchExpansionTop1Scorer)
+    worker.scorer.score_proposals.side_effect = ValueError(exception_secret)
+
+    # Create batch with combination of terminal/non-terminal prefill chunks
+    # and decodes (different seq_ids).
+    decodes, _, _ = create_batch(batch_size, k)
+    # Pre-chunking here, get 'batch_size' chunks.
+    prefill, _, _ = create_batch(batch_size,
+                                 k,
+                                 prefill_chunk_size=4,
+                                 seq_ids=list(range(batch_size,
+                                                    batch_size * 2)))
+
+    if batch_composition == "prefill_only":
+        n_prefills = batch_size
+    elif batch_composition == "decode_only":
+        n_prefills = 0
+    else:
+        n_prefills = random.randint(1, batch_size - 1)
+    n_decodes = batch_size - n_prefills
+
+    prefill = random.sample(prefill, n_prefills)
+    decodes = random.sample(decodes, n_decodes)
+    target_group_metadata_list = prefill + decodes
+    execute_model_req = ExecuteModelRequest(
+        seq_group_metadata_list=target_group_metadata_list,
+        num_lookahead_slots=k)
+
+    target_token_ids = torch.randint(low=0,
+                                     high=vocab_size,
+                                     size=(1, batch_size * (k + 1)),
+                                     dtype=torch.int64,
+                                     device='cuda')
+    target_token_probs = torch.rand(1,
+                                    batch_size * (k + 1),
+                                    vocab_size,
+                                    dtype=torch.float32,
+                                    device='cuda')
+    target_token_logprobs = torch.rand(1,
+                                       batch_size * (k + 1),
+                                       vocab_size,
+                                       dtype=torch.float32,
+                                       device='cuda')
+    target_output = create_sampler_output_list(target_token_ids,
+                                               target_token_probs,
+                                               target_token_logprobs)
+
+    target_worker.execute_model.return_value = [target_output[0]]
+
+    if not len(decodes):
+        worker.execute_model(execute_model_req=execute_model_req)
+        # no spec run (prefill only)
+        draft_worker.execute_model.assert_called_once_with(execute_model_req)
+        target_worker.execute_model.assert_called_once_with(execute_model_req)
+    else:
+        # Decode-only run OR mixed batch, scorer call fails (it's mocked)
+        with pytest.raises(ValueError, match=exception_secret):
+            worker.execute_model(execute_model_req=execute_model_req)
+        # but first draft still counted
+        assert draft_worker.get_spec_proposals.call_count == 1

tests/spec_decode/utils.py

@@ -146,6 +146,41 @@ def create_seq_group_metadata_from_prompts(
     return seq_grou_metadata_list
 
 
+def create_chunked_seq_group_metadata_from_prompt(
+        prompt: List[int],
+        num_gpu_blocks: int,
+        chunk_size: int,
+        block_size: int,
+        seq_id: Optional[int] = None) -> List[SequenceGroupMetadata]:
+
+    if seq_id is None:
+        seq_id = 0
+
+    free_gpu_blocks = list(range(num_gpu_blocks))
+
+    block_allocations = [
+        free_gpu_blocks.pop()
+        for _ in range(round_up_to_next_block(len(prompt), block_size))
+    ]
+
+    seq_group_metadata_list = []
+    for i, idx in enumerate(range(0, len(prompt), chunk_size)):
+        chunk_ids = prompt[idx:idx + chunk_size]
+        data = SequenceData.from_seqs(prompt)
+        data.update_num_computed_tokens(idx)
+        seq_data = {i: data}
+        seq_group_metadata_list.append(
+            SequenceGroupMetadata(
+                request_id=str(seq_id),
+                is_prompt=True,
+                do_sample=idx + chunk_size >= len(prompt),  # terminal chunk
+                seq_data=seq_data,
+                sampling_params=SamplingParams(temperature=0.0),
+                block_tables={i: block_allocations},
+                token_chunk_size=len(chunk_ids)))
+    return seq_group_metadata_list
+
+
 def assert_logprobs_dict_allclose(
         actual_logprobs: List[Dict[int, Logprob]],
         expected_logprobs: List[Dict[int, Logprob]]) -> None:
@@ -198,7 +233,8 @@ def create_batch(batch_size,
                  prev_output_token_len: int = 10,
                  seq_ids: Optional[List[int]] = None,
                  num_gpu_blocks: Optional[int] = None,
-                 block_size: Optional[int] = None):
+                 block_size: Optional[int] = None,
+                 prefill_chunk_size: Optional[int] = None):
     if block_size is None:
         block_size = 8
 
@@ -213,6 +249,19 @@ def create_batch(batch_size,
         prompt_lens = prompt_len
 
     prompts = [[next(iterator) for _ in range(p_len)] for p_len in prompt_lens]
+
+    if prefill_chunk_size:
+        # Create a batch of chunked prompts.
+        if not seq_ids:
+            seq_ids = list(range(len(prompts)))
+        seq_group_metadata_list = []
+        for p, sid in zip(prompts, seq_ids):
+            seq_group_metadata_list += \
+                create_chunked_seq_group_metadata_from_prompt(
+                p, num_gpu_blocks, prefill_chunk_size, block_size, sid)
+        seq_group_metadata_list = seq_group_metadata_list[:batch_size]
+        prev_output_tokens = []
+    else:
         prev_output_tokens = [[
             next(iterator) for _ in range(prev_output_token_len)
         ] for _ in range(batch_size)]

vllm/attention/backends/flash_attn.py

@@ -276,7 +276,11 @@ class FlashAttentionMetadata(AttentionMetadata):
             max_query_len=self.max_query_len,
             max_prefill_seq_len=0,
             max_decode_seq_len=self.max_decode_seq_len,
-            query_start_loc=self.query_start_loc[self.num_prefills:]
+            # Batch may be composed of prefill|decodes, adjust query start
+            # indices to refer to the start of decodes. E.g.
+            # in tokens:[3 prefills|6 decodes], query_start_loc=[3,9] => [0,6].
+            query_start_loc=(self.query_start_loc[self.num_prefills:] -
+                             self.query_start_loc[self.num_prefills])
             if self.query_start_loc is not None else None,
             seq_start_loc=self.seq_start_loc[self.num_prefills:]
             if self.seq_start_loc is not None else None,
@@ -903,7 +907,9 @@ def unified_flash_attention(
         # Decoding run.
         # Use flash_attn_varlen_func kernel for speculative decoding
         # because different queries might have different lengths.
+
         assert decode_meta.max_decode_query_len is not None
+        # use only for actual varlen decoding
         if decode_meta.max_decode_query_len > 1:
             assert attn_type == AttentionType.DECODER, (
                 "Only decoder-only models support max_decode_query_len > 1")
@@ -949,8 +955,6 @@ def unified_flash_attention(
         assert prefill_output is not None
         return prefill_output.view(num_prefill_query_tokens, hidden_size)
 
-    # Chunked prefill does not work with speculative decoding.
-    # Therefore, the query length for decode should be 1 in chunked prefill.
     assert decode_meta is not None
     decode_output = decode_output.squeeze(1)
     output = torch.cat([prefill_output, decode_output], dim=0)

vllm/attention/backends/rocm_flash_attn.py

@@ -192,6 +192,12 @@ class ROCmFlashAttentionMetadata(AttentionMetadata, PagedAttentionMetadata):
             block_tables=self.block_tables[self.num_prefills:],
             use_cuda_graph=self.use_cuda_graph,
         )
+        # Batch may be composed of prefill|decodes, adjust query start indices
+        # to refer to the start of decodes when the two are split apart.
+        # E.g. in tokens:[3 prefills|6 decodes], query_start_loc=[3,9] => [0,6].
+        if self._cached_decode_metadata.query_start_loc is not None:
+            qs = self._cached_decode_metadata.query_start_loc
+            self._cached_decode_metadata.query_start_loc = qs - qs[0]
         return self._cached_decode_metadata
 
     def advance_step(self,

vllm/attention/backends/xformers.py

@@ -272,6 +272,13 @@ class XFormersMetadata(AttentionMetadata, PagedAttentionMetadata):
             max_encoder_seq_len=self.max_encoder_seq_len,
             cross_slot_mapping=self.cross_slot_mapping,
             cross_block_tables=self.cross_block_tables)
+
+        # Batch may be composed of prefill|decodes, adjust query start indices
+        # to refer to the start of decodes when the two are split apart.
+        # E.g. in tokens:[3 prefills|6 decodes], query_start_loc=[3,9] => [0,6].
+        if self._cached_decode_metadata.query_start_loc is not None:
+            qs = self._cached_decode_metadata.query_start_loc
+            self._cached_decode_metadata.query_start_loc = qs - qs[0]
         return self._cached_decode_metadata
 
 

vllm/config.py

@@ -192,7 +192,6 @@ class ModelConfig:
         self.max_logprobs = max_logprobs
         self.disable_sliding_window = disable_sliding_window
         self.skip_tokenizer_init = skip_tokenizer_init
-
         self.hf_config = get_config(self.model, trust_remote_code, revision,
                                     code_revision, rope_scaling, rope_theta,
                                     config_format)
@@ -1317,13 +1316,6 @@ class SpeculativeConfig:
                              "speculative decoding is > 1, but got "
                              f"{speculative_disable_by_batch_size=}")
 
-        # Reminder: Please update docs/source/serving/compatibility_matrix.rst
-        # If the feature combo become valid
-        if enable_chunked_prefill:
-            raise ValueError(
-                "Speculative decoding and chunked prefill are "
-                f"currently mutually exclusive ({enable_chunked_prefill=}).")
-
         # TODO: The user should be able to specify revision/max model len
         # for the draft model. It is not currently supported.
         draft_revision = None
@@ -1390,6 +1382,12 @@ class SpeculativeConfig:
                         f"num_speculative_tokens={n_predict}, but "
                         f"{num_speculative_tokens=} was provided.")
 
+            if enable_chunked_prefill and draft_hf_config.model_type in (
+                    "medusa", "mlp_speculator", "eagle"):
+                raise ValueError(
+                    "Chunked prefill and hidden-state based draft models are "
+                    "not compatible.")
+
             draft_model_config.max_model_len = (
                 SpeculativeConfig._maybe_override_draft_max_model_len(
                     speculative_max_model_len,

vllm/core/scheduler.py

@@ -1147,6 +1147,7 @@ class Scheduler:
 
         # Update swapped requests.
         self.swapped.extend(running_scheduled.swapped_out)
+        # Put prefills first due to Attention backend ordering assumption.
         return SchedulerOutputs(
             scheduled_seq_groups=(prefills.seq_groups +
                                   running_scheduled.prefill_seq_groups +

vllm/engine/output_processor/multi_step.py

@@ -134,8 +134,10 @@ class MultiStepOutputProcessor(SequenceGroupOutputProcessor):
                 sample for sample in samples
                 if sample.output_token != VLLM_INVALID_TOKEN_ID
             ]
-            assert valid_samples
 
+            # When both spec-decode and pre-fill chunking are enabled, we
+            # don't have guaranteed samples here (e.g. all -1s).
+            if valid_samples:
                 self._process_seq_outputs(seq, valid_samples,
                                           sequence_group.sampling_params)
 

vllm/spec_decode/batch_expansion.py

@@ -90,7 +90,7 @@ class BatchExpansionTop1Scorer(SpeculativeScorer):
         else:
             # Batch has a mix of spec decode enabled and disabled seq groups
             contracted = self._contract_batch(
-                contracted_bs=len(execute_model_req.seq_group_metadata_list),
+                execute_model_req.seq_group_metadata_list,
                 target_sampler_output=target_sampler_output,
                 proposals=proposals,
                 num_scoring_tokens=num_scoring_tokens,
@@ -126,7 +126,7 @@ class BatchExpansionTop1Scorer(SpeculativeScorer):
             split_batch_by_proposal_len(
                 seq_group_metadata_list, proposal_lens_list)
 
-        target_seq_group_metadata_list = self._create_scoring_model_input(
+        spec_expanded_seqs = self._create_scoring_model_input(
             seq_group_metadata_list=spec_seqs,
             proposal_token_ids=proposal_token_ids_list,
             # NOTE: We determine the seq ids in the expanded batch using the
@@ -135,16 +135,19 @@ class BatchExpansionTop1Scorer(SpeculativeScorer):
                 seq_ids=get_all_seq_ids(seq_group_metadata_list)),
         )
 
-        num_scoring_tokens = len(target_seq_group_metadata_list)
-        target_seq_group_metadata_list.extend(non_spec_seqs)
+        num_scoring_tokens = len(spec_expanded_seqs)
+        # Batch speculative and non-speculative (e.g. chunked prefill) requests
+        # but make sure order is prefill|decode due to backend requirement.
+        target_seq_group_metadata_list = non_spec_seqs + spec_expanded_seqs
 
         return (spec_indices, non_spec_indices, target_seq_group_metadata_list,
                 num_scoring_tokens)
 
     def _contract_batch(
-        self, contracted_bs: int, target_sampler_output: SamplerOutput,
-        proposals: SpeculativeProposals, num_scoring_tokens: int,
-        non_spec_indices: List[int], spec_indices: List[int], k: int
+        self, contracted_seq_group_metadata_list: List[SequenceGroupMetadata],
+        target_sampler_output: SamplerOutput, proposals: SpeculativeProposals,
+        num_scoring_tokens: int, non_spec_indices: List[int],
+        spec_indices: List[int], k: int
     ) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor,
                Optional[torch.Tensor]]:
         """Contract the expanded batch back into its original size.
@@ -154,6 +157,7 @@ class BatchExpansionTop1Scorer(SpeculativeScorer):
         contracted_bs is the original batch size, and the batch size that the
         target_sampler_output will be contracted to.
         """
+        contracted_bs = len(contracted_seq_group_metadata_list)
         (target_token_ids, target_probs, target_logprobs, target_hidden_states,
          non_spec_target_token_ids, non_spec_target_probs,
          non_spec_target_logprobs,
@@ -166,8 +170,8 @@ class BatchExpansionTop1Scorer(SpeculativeScorer):
 
         # The number of tokens in the expanded batch used for speculation is
         # equal to the total expanded batch size minus the number of samples for
-        # non-speculative sequences.
-        non_spec_expanded_bs = len(non_spec_target_token_ids)
+        # non-speculative sequences, prefill chunks with no out tokens included
+        non_spec_expanded_bs = len(non_spec_indices)
         spec_expanded_bs = expanded_batch_size - non_spec_expanded_bs
 
         target_token_ids = target_token_ids.reshape(spec_expanded_bs, k + 1)
@@ -191,7 +195,12 @@ class BatchExpansionTop1Scorer(SpeculativeScorer):
         else:
             all_hidden_states = None
 
-        if non_spec_indices:
+        # Rule out prefills that produce no tokens.
+        non_spec_indices = [
+            idx for idx in non_spec_indices
+            if contracted_seq_group_metadata_list[idx].do_sample
+        ]
+        if len(non_spec_indices):
             all_tokens[non_spec_indices, :1] = \
                 non_spec_target_token_ids.unsqueeze(1)
             all_probs[non_spec_indices, :1, :] = \
@@ -290,9 +299,6 @@ class BatchExpansionTop1Scorer(SpeculativeScorer):
         This function creates K+1 target SequenceGroupMetadata to take
         advantage of the bonus token.
         """
-        assert not input_seq_group_metadata.is_prompt, (
-            "Speculating on "
-            "prompts not yet supported")
         assert len(input_seq_group_metadata.seq_data) == 1, (
             "Beam search "
             "not supported in speculative decoding")
@@ -390,27 +396,22 @@ class BatchExpansionTop1Scorer(SpeculativeScorer):
         # and non spec sequences) and should be removed in the future. It can be
         # done by supporting per-sequence proposal lens.
         #
-        # First samples are from speculative scoring, latter samples are non-
-        # speculative samples.
-        split_sizes = (num_scoring_tokens,
-                       sampler_output.sampled_token_ids.numel() -
-                       num_scoring_tokens)
-        (spec_probs, non_spec_probs
-         ) = sampler_output.sampled_token_probs.split(split_sizes)
-        (spec_sampled_tokens, non_spec_sampled_tokens
+        # First samples are non-speculative, latter samples are from speculative
+        # scoring (prefill|decode order).
+        split_sizes = (sampler_output.sampled_token_ids.numel() -
+                       num_scoring_tokens, num_scoring_tokens)
+        (non_spec_probs,
+         spec_probs) = sampler_output.sampled_token_probs.split(split_sizes)
+        (non_spec_sampled_tokens, spec_sampled_tokens
          ) = sampler_output.sampled_token_ids.flatten().split(split_sizes)
-        (
-            spec_logprobs,
-            non_spec_logprobs,
-        ) = sampler_output.logprobs.split(split_sizes)
+        (non_spec_logprobs,
+         spec_logprobs) = sampler_output.logprobs.split(split_sizes)
 
         if sampler_output.hidden_states is not None:
-            (
-                spec_hidden_states,
-                non_spec_hidden_states,
+            (non_spec_hidden_states, spec_hidden_states
              ) = sampler_output.hidden_states.split(split_sizes)
         else:
-            spec_hidden_states, non_spec_hidden_states = None, None
+            non_spec_hidden_states, spec_hidden_states = None, None
 
         return (spec_sampled_tokens, spec_probs, spec_logprobs,
                 spec_hidden_states, non_spec_sampled_tokens, non_spec_probs,

vllm/spec_decode/mqa_scorer.py

@@ -21,6 +21,11 @@ class MQAScorer(SpeculativeScorer):
         all_proposal_lengths = proposals.proposal_lens.tolist()
         for i, seq_group_metadata in enumerate(
                 execute_model_req.seq_group_metadata_list):
+            if all_proposal_lengths[i] == 0:
+                # Keep prompt seqs untouched (keep computed_tokens for chunks).
+                target_seq_group_metadata_list.append(seq_group_metadata)
+                continue
+
             seq_data_dict = seq_group_metadata.seq_data
             assert len(seq_data_dict) == 1
             seq_id = next(iter(seq_data_dict.keys()))
@@ -40,8 +45,7 @@ class MQAScorer(SpeculativeScorer):
             new_seq_data.update_num_computed_tokens(
                 len(prompt_token_ids) + len(output_token_ids) - 1)
 
-            # Ensure that the new sequence has at least one token
-            # because we only use mqa scorer in the decoding stage.
+            # Ensure that the new decode sequence has at least one token.
             assert len(output_token_ids) >= 1
             new_seq_data_dict = {target_seq_id: new_seq_data}
 
@@ -54,7 +58,6 @@ class MQAScorer(SpeculativeScorer):
                     target_seq_id: seq_group_metadata.block_tables[seq_id],
                 },
                 lora_request=None,
-                token_chunk_size=1,
             )
             target_seq_group_metadata_list.append(new_seq_group_metadata)
 
@@ -77,6 +80,7 @@ class MQAScorer(SpeculativeScorer):
             all_probs = target_probs.reshape(bs, k + 1, self._vocab_size)
             all_logprobs = target_logprobs.reshape(bs, k + 1, self._vocab_size)
         else:
+            # We either have decodes with different lens or prefill+decodes.
             all_tokens = target_token_ids.new_full(size=(bs, k + 1),
                                                    fill_value=-1)
             all_probs = target_probs.new_zeros(*all_tokens.shape,
@@ -85,7 +89,10 @@ class MQAScorer(SpeculativeScorer):
                                                     fill_value=-float("inf"))
             target_token_ids = target_token_ids.flatten()
             start_loc = 0
-            for i, proposed_len in enumerate(all_proposal_lengths):
+            for i, (proposed_len, seq_meta) in enumerate(
+                    zip(all_proposal_lengths, target_seq_group_metadata_list)):
+                # Skip chunks with no output tokens.
+                if seq_meta.do_sample:
                     output_len = proposed_len + 1
                     end_loc = start_loc + output_len
                     all_tokens[

vllm/spec_decode/spec_decode_worker.py

@@ -418,7 +418,10 @@ class SpecDecodeWorker(LoraNotSupportedWorkerBase):
         #    none of the requests in the batch have spec decoding enabled.
         # In any of these cases, the proposer and scorer workers
         # are called normally.
-        no_spec = num_lookahead_slots == 0 or disable_all_speculation or all(
+        # We expect `num_speculative_tokens` to be None for prefills.
+        no_spec = all(
+            sgm.is_prompt for sgm in execute_model_req.seq_group_metadata_list
+        ) or num_lookahead_slots == 0 or disable_all_speculation or all(
             sgm.num_speculative_tokens == 0
             for sgm in execute_model_req.seq_group_metadata_list)
 
@@ -484,7 +487,7 @@ class SpecDecodeWorker(LoraNotSupportedWorkerBase):
 
     def _serialize_sampler_output_no_logprobs(
             self, execute_model_req: ExecuteModelRequest,
-            sampler_output: SamplerOutput) -> SamplerOutput:
+            sampler_output: SamplerOutput) -> List[SamplerOutput]:
         """
         Creates and returns a `SamplerOutput` with only the token IDs being
         serialized to CPU and populated in `CompletionSequenceGroupOutput`.
@@ -514,15 +517,29 @@ class SpecDecodeWorker(LoraNotSupportedWorkerBase):
             if any(seq_output_prompt_logprobs) else \
                 sampler_output.sampled_token_ids).tolist()
 
-        seq_data_entries = (
+        seq_data_entries = [
             (seq_id, seq_data) for sg in \
             execute_model_req.seq_group_metadata_list \
             for seq_id, seq_data in sg.seq_data.items()
-        )
+            if sg.do_sample # ignore empty token sequences
+        ]
         completion_seq_group_output_list: List[
             CompletionSequenceGroupOutput] = []
-        for index, ((seq_id, seq_data), needs_prompt_logprobs) in \
-            enumerate(zip(seq_data_entries, seq_output_prompt_logprobs)):
+        output_index = 0
+        # Make sure the non-terminal prefill chunks are still aligned with
+        # their own empty output.
+        for seq_group_meta in execute_model_req.seq_group_metadata_list:
+            # Since we can get chunks here, we dont always have a sampled token
+            # (only on last chunk) but we still have to provide an output.
+            if not seq_group_meta.do_sample:
+                completion_seq_group_output_list.append(
+                    CompletionSequenceGroupOutput(samples=[],
+                                                  prompt_logprobs=None))
+            else:
+                # Sequence with output.
+                seq_id, seq_data = seq_data_entries[output_index]
+                needs_prompt_logprobs = seq_output_prompt_logprobs[
+                    output_index]
                 if needs_prompt_logprobs:
                     prompt_token_ids = seq_data.get_prompt_token_ids()
                     prompt_logprobs = [
@@ -538,17 +555,18 @@ class SpecDecodeWorker(LoraNotSupportedWorkerBase):
                     ]
                 else:
                     prompt_logprobs = None
-
                 completion_seq_group_output_list.append(
                     create_sequence_group_output(
-                    token_id=sampled_token_ids_list[index][0],
+                        token_id=sampled_token_ids_list[output_index][0],
                         token_id_logprob_rank=-1,
                         token_id_logprob=0.0,
                         seq_id=seq_id,
                         topk_token_ids=[],
                         topk_logprobs=[],
                         prompt_logprobs=prompt_logprobs))
-        return SamplerOutput(outputs=completion_seq_group_output_list)
+                output_index += 1
+
+        return [SamplerOutput(outputs=completion_seq_group_output_list)]
 
     @nvtx_range("spec_decode_worker._run_no_spec")
     def _run_no_spec(self, execute_model_req: ExecuteModelRequest,
@@ -568,6 +586,9 @@ class SpecDecodeWorker(LoraNotSupportedWorkerBase):
         hidden_states = sampler_output.hidden_states
         if hidden_states is not None:
             # remove hidden_states for prompt tokens
+            # TODO Enable `return_hidden_states`: prefill chunks hidden states
+            # are pruned by the logits processor. Also, they should be arranged
+            # back into full-prefill latent. Address it to enable MLPSpeculator.
             if any(seq.is_prompt
                    for seq in execute_model_req.seq_group_metadata_list):
                 hidden_states = hidden_states[
@@ -593,14 +614,14 @@ class SpecDecodeWorker(LoraNotSupportedWorkerBase):
         sampler_output_to_return = (self._serialize_sampler_output_no_logprobs(
             execute_model_req=execute_model_req, sampler_output=sampler_output)
                                     if self._disable_logprobs else
-                                    sampler_output)
+                                    [sampler_output])
 
         # Clear device tensors from sampler output. This reduces communication
         # overhead when the engine runs in a different process than the workers.
         sampler_output.sampled_token_probs = None
         sampler_output.sampled_token_ids = None
         sampler_output.logprobs = None
-        return [sampler_output_to_return]
+        return sampler_output_to_return
 
     def _run_non_driver_rank(self) -> bool:
         """Run proposer and verifier model in non-driver workers. This is used
@@ -644,9 +665,15 @@ class SpecDecodeWorker(LoraNotSupportedWorkerBase):
         This invokes the proposer worker to get k speculative tokens for each
         sequence, then scores each speculative token using the scoring worker.
 
+        When `enable_chunked_prefill` is set, scorer will batch decodes and 
+        prefills, while proposer will sync its KV-cache by running an extra
+        forward on prefills.
+
         Returns a list of SamplerOutput, each containing a single token per
         sequence.
         """
+        # With prefill chunking, expect requests to have prompts first
+        # so that backend gets prefill|decode.
         assert num_lookahead_slots == execute_model_req.num_lookahead_slots
 
         # Pass last hidden states from target model to proposer
@@ -671,6 +698,25 @@ class SpecDecodeWorker(LoraNotSupportedWorkerBase):
                 proposals,
             )
 
+        _, (non_spec_seqs, non_spec_indices) = split_batch_by_proposal_len(
+            execute_model_req.seq_group_metadata_list, proposals.proposal_lens)
+        # With prefill chunking enabled, `non_spec_seqs` contains prefills too:
+        # discard decodes that have already been processed by proposer.
+        non_spec_indices = [
+            idx for idx in non_spec_indices
+            if execute_model_req.seq_group_metadata_list[idx].is_prompt
+        ]
+        if len(non_spec_indices):
+            all_hidden_states = proposal_scores.hidden_states
+            # TODO fix `return_hidden_states`, same as in `_run_no_spec`
+            if all_hidden_states is not None:
+                prefill_hidden_states = all_hidden_states[non_spec_indices]
+                execute_model_req.previous_hidden_states = \
+                    prepare_prefill_hidden_states(prefill_hidden_states)
+            # Sync proposer KV cache for prefills.
+            prefill_req = execute_model_req.clone(non_spec_seqs)
+            self.proposer_worker.execute_model(prefill_req)
+
         with Timer() as verification_timer:
             accepted_token_ids, target_logprobs = self._verify_tokens(
                 execute_model_req.seq_group_metadata_list, proposal_scores,
@@ -769,7 +815,6 @@ class SpecDecodeWorker(LoraNotSupportedWorkerBase):
             self.previous_hidden_states = HiddenStates(
                 hidden_states, seq_group_metadata_list,
                 second_last_token_hidden_states)
-
         return accepted_token_ids, logprobs
 
     def _create_output_sampler_list(
@@ -819,6 +864,8 @@ class SpecDecodeWorker(LoraNotSupportedWorkerBase):
         accepted_token_ids_by_step = accepted_token_ids_by_step.tolist()
 
         # Construct the output on a per-step, per-sequence basis.
+        # Non-terminal prefill chunks will end up here as rows with just -1s
+        # i.e mixed-batch [[-1, 1576], [-1, 29884], [-1, -1], [-1, -1]]
         sampler_output_list: List[SamplerOutput] = []
         for step_index in range(num_steps):
             if all(token_id == -1
@@ -861,7 +908,6 @@ class SpecDecodeWorker(LoraNotSupportedWorkerBase):
             # This is periodic because the rejection sampler emits metrics
             # periodically.
             self._maybe_log_stage_times(*stage_times)
-
         return sampler_output_list
 
     def _maybe_log_stage_times(self, average_time_per_proposal_tok_ms: float,

vllm/spec_decode/top1_proposer.py

@@ -109,7 +109,6 @@ class Top1Proposer(SpeculativeProposer):
             proposal_probs=proposal_probs,
             proposal_lens=proposal_lens,
             no_proposals=maybe_sampler_output is None)
-
         return proposals
 
     def _split_by_proposal_len(
@@ -127,9 +126,10 @@ class Top1Proposer(SpeculativeProposer):
         nonzero_proposal_len_seqs: List[SequenceGroupMetadata] = []
         nonzero_proposal_len_indices: List[int] = []
         for i, seq_group_metadata in enumerate(seq_group_metadata_list):
-            # The speculative decoding for this request has been disabled
-            # (e.g. due to high traffic).
-            if seq_group_metadata.num_speculative_tokens == 0:
+            # The speculative decoding for this request has either been disabled
+            # (e.g. due to high traffic) or this is a prompt request.
+            if (seq_group_metadata.is_prompt
+                    or seq_group_metadata.num_speculative_tokens == 0):
                 proposal_lens.append(0)
                 continue