[Speculative Decoding] Fixing hidden states handling in batch expansion (#7508)

tests/spec_decode/e2e/conftest.py

@@ -288,7 +288,8 @@ def run_greedy_equality_correctness_test(baseline_llm_generator,
                                   ensure_all_accepted=ensure_all_accepted)
 
 
-def run_equality_correctness_test(baseline_llm_generator,
+def run_equality_correctness_test(
+        baseline_llm_generator,
         test_llm_generator,
         batch_size,
         max_output_len,
@@ -296,7 +297,8 @@ def run_equality_correctness_test(baseline_llm_generator,
         temperature: float,
         seeded: bool,
         print_tokens: bool = False,
-                                  ensure_all_accepted: bool = False):
+        ensure_all_accepted: bool = False,
+        expected_acceptance_rate: Optional[float] = None):
     """Helper method that compares the outputs of both the baseline LLM and
     the test LLM. It asserts greedy equality, e.g. that the outputs are exactly
     the same when temperature is zero (or when temperature is > 0 and seeded).
@@ -357,5 +359,10 @@ def run_equality_correctness_test(baseline_llm_generator,
         print(f'{i=}     {spec_token_ids=}')
         assert baseline_token_ids == spec_token_ids
 
+    print(f'{acceptance_rate=}')
+
     if ensure_all_accepted:
         assert acceptance_rate == 1.0
+
+    if expected_acceptance_rate is not None:
+        assert acceptance_rate >= expected_acceptance_rate - 1e-2

tests/spec_decode/e2e/test_mlp_correctness.py

@@ -82,6 +82,48 @@ def test_mlp_e2e_greedy_correctness(baseline_llm_generator, test_llm_generator,
                                          force_output_len=True)
 
 
+@pytest.mark.parametrize(
+    "common_llm_kwargs",
+    [{
+        # Skip cuda graph recording for fast test.
+        "enforce_eager": True,
+
+        # Required for spec decode.
+        "use_v2_block_manager": True,
+
+        # Print spec metrics.
+        "disable_log_stats": False,
+
+        # Precision
+        "dtype": PRECISION,
+
+        # Main model
+        "model": MAIN_MODEL,
+    }])
+@pytest.mark.parametrize("per_test_common_llm_kwargs", [{}])
+@pytest.mark.parametrize("baseline_llm_kwargs", [{}])
+@pytest.mark.parametrize("test_llm_kwargs", [
+    {
+        "speculative_model": SPEC_MODEL,
+    },
+])
+@pytest.mark.parametrize("output_len", [2048])
+@pytest.mark.parametrize("batch_size", [1, 32])
+@pytest.mark.parametrize("seed", [1])
+def test_mlp_e2e_acceptance_rate(baseline_llm_generator, test_llm_generator,
+                                 batch_size: int, output_len: int):
+    """Verify acceptance rate with different batch size and large output 
+    length."""
+    run_equality_correctness_test(baseline_llm_generator,
+                                  test_llm_generator,
+                                  batch_size,
+                                  max_output_len=output_len,
+                                  temperature=0.0,
+                                  seeded=True,
+                                  force_output_len=True,
+                                  expected_acceptance_rate=0.48)
+
+
 @pytest.mark.parametrize(
     "common_llm_kwargs",
     [{

vllm/spec_decode/batch_expansion.py

@@ -1,6 +1,6 @@
 from array import array
 from itertools import chain, count
-from typing import Iterator, List, Tuple
+from typing import Iterator, List, Optional, Tuple
 
 import torch
 
@@ -88,7 +88,8 @@ class BatchExpansionTop1Scorer(SpeculativeScorer):
         assert len(target_sampler_output) == 1, "expected single-step output"
         target_sampler_output = target_sampler_output[0]
 
-        all_tokens, all_probs, spec_logprobs = self._contract_batch(
+        (all_tokens, all_probs, spec_logprobs,
+         all_hidden_states) = self._contract_batch(
              contracted_bs=len(execute_model_req.seq_group_metadata_list),
              target_sampler_output=target_sampler_output,
              proposals=proposals,
@@ -102,7 +103,7 @@ class BatchExpansionTop1Scorer(SpeculativeScorer):
             probs=all_probs,
             token_ids=all_tokens,
             logprobs=spec_logprobs,
-            hidden_states=target_sampler_output.hidden_states,
+            hidden_states=all_hidden_states,
         )
 
     def _expand_batch(
@@ -147,8 +148,9 @@ class BatchExpansionTop1Scorer(SpeculativeScorer):
     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],
+        non_spec_indices: List[int], spec_indices: List[int], k: int
-            k: int) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor]:
+    ) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor,
+               Optional[torch.Tensor]]:
         """Contract the expanded batch back into its original size.
         This maps the scores of speculative tokens back to their original
         sequences.
@@ -156,9 +158,10 @@ class BatchExpansionTop1Scorer(SpeculativeScorer):
         contracted_bs is the original batch size, and the batch size that the
         target_sampler_output will be contracted to.
         """
-        (target_token_ids, target_probs, target_logprobs,
+        (target_token_ids, target_probs, target_logprobs, target_hidden_states,
          non_spec_target_token_ids, non_spec_target_probs,
-         non_spec_target_logprobs) = self._split_scoring_output(
+         non_spec_target_logprobs,
+         non_spec_target_hidden_states) = self._split_scoring_output(
              target_sampler_output, num_scoring_tokens)
 
         # Map distinct sequences used to score each token
@@ -176,23 +179,40 @@ class BatchExpansionTop1Scorer(SpeculativeScorer):
                                             self._vocab_size)
         target_logprobs = target_logprobs.reshape(target_probs.shape)
 
+        if target_hidden_states is not None:
+            target_hidden_states = target_hidden_states.reshape(
+                spec_expanded_bs, k + 1, target_hidden_states.shape[-1])
+
         all_tokens = target_token_ids.new_full(size=(contracted_bs, k + 1),
                                                fill_value=-1)
         all_probs = target_probs.new_zeros(*all_tokens.shape, self._vocab_size)
         all_logprobs = target_logprobs.new_full(size=all_probs.shape,
                                                 fill_value=-float("inf"))
 
+        if target_sampler_output.hidden_states is not None:
+            all_hidden_states = target_hidden_states.new_zeros(
+                size=(contracted_bs, k + 1, target_hidden_states.shape[-1]))
+        else:
+            all_hidden_states = None
+
         if non_spec_indices:
             all_tokens[non_spec_indices, :1] = non_spec_target_token_ids
             all_probs[non_spec_indices, :1, :] = non_spec_target_probs
             all_logprobs[non_spec_indices, :1, :] = non_spec_target_logprobs
 
+            if all_hidden_states is not None:
+                all_hidden_states[
+                    non_spec_indices, :1, :] = non_spec_target_hidden_states
+
         if spec_indices:
             all_tokens[spec_indices] = target_token_ids
             all_probs[spec_indices] = target_probs
             all_logprobs[spec_indices] = target_logprobs
 
-        return all_tokens, all_probs, all_logprobs
+            if all_hidden_states is not None:
+                all_hidden_states[spec_indices] = target_hidden_states
+
+        return all_tokens, all_probs, all_logprobs, all_hidden_states
 
     def _create_scoring_model_input(
         self,
@@ -327,8 +347,9 @@ class BatchExpansionTop1Scorer(SpeculativeScorer):
 
     def _split_scoring_output(
         self, sampler_output: SamplerOutput, num_scoring_tokens: int
-    ) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor,
+    ) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor,
-               torch.Tensor, torch.Tensor]:
+               Optional[torch.Tensor], torch.Tensor, torch.Tensor,
+               torch.Tensor, Optional[torch.Tensor]]:
         """Split the target model output into speculative and non-speculative
         output.
         """
@@ -353,24 +374,37 @@ class BatchExpansionTop1Scorer(SpeculativeScorer):
             non_spec_logprobs,
         ) = sampler_output.logprobs.split(split_sizes)
 
+        if sampler_output.hidden_states is not None:
+            (
+                spec_hidden_states,
+                non_spec_hidden_states,
+            ) = sampler_output.hidden_states.split(split_sizes)
+        else:
+            spec_hidden_states, non_spec_hidden_states = None, None
+
         # Convert scores to tensors.
         sampler_output.sampled_token_probs = spec_probs
         sampler_output.sampled_token_ids = spec_sampled_tokens
         sampler_output.logprobs = spec_logprobs
-        (target_token_ids, target_probs,
+        sampler_output.hidden_states = spec_hidden_states
-         target_logprobs) = sampler_output_to_torch([sampler_output], True)
+        (target_token_ids, target_probs, target_logprobs,
+         target_hidden_states) = sampler_output_to_torch([sampler_output],
+                                                         True)
 
         # Convert non-speculative output tokens to tensors.
         sampler_output.sampled_token_probs = non_spec_probs
         sampler_output.sampled_token_ids = non_spec_sampled_tokens
         sampler_output.logprobs = non_spec_logprobs
+        sampler_output.hidden_states = non_spec_hidden_states
         (non_spec_target_token_ids, non_spec_target_probs,
-         non_spec_target_logprobs) = sampler_output_to_torch([sampler_output],
+         non_spec_target_logprobs,
-                                                             True)
+         non_spec_target_hidden_states) = sampler_output_to_torch(
+             [sampler_output], True)
 
         return (target_token_ids, target_probs, target_logprobs,
-                non_spec_target_token_ids, non_spec_target_probs,
+                target_hidden_states, non_spec_target_token_ids,
-                non_spec_target_logprobs)
+                non_spec_target_probs, non_spec_target_logprobs,
+                non_spec_target_hidden_states)
 
     def _create_target_seq_id_iterator(
             self, seq_ids: List[SeqId]) -> Iterator[TargetSeqId]:

vllm/spec_decode/spec_decode_worker.py

@@ -646,9 +646,8 @@ class SpecDecodeWorker(LoraNotSupportedWorkerBase):
         hidden_states = proposal_scores.hidden_states
         if hidden_states is not None:
             # Contract hidden states based on accepted tokens
-            hs_size = hidden_states.shape[1]
+            hs_size = hidden_states.shape[-1]
-            hidden_states = hidden_states.reshape(-1, max_proposal_len + 1,
+
-                                                  hs_size)
             accepted_index = accepted_token_ids + 1  # Convert -1 to 0
             accepted_index = accepted_index.count_nonzero(dim=1).add_(-1)
             index = accepted_index[:, None, None].expand(-1, 1, hs_size)

vllm/spec_decode/top1_proposer.py

@@ -242,7 +242,7 @@ class Top1Proposer(SpeculativeProposer):
             return proposal_tokens, proposal_probs, proposal_lens_tensor
 
         sampler_output = maybe_sampler_output
-        proposal_tokens, proposal_probs, _ = sampler_output_to_torch(
+        proposal_tokens, proposal_probs, *_ = sampler_output_to_torch(
             sampler_output, sampler_transposed)
 
         # Now, reformat the output GPU tensors such that each sequence has

vllm/spec_decode/util.py

@@ -123,7 +123,7 @@ def split_batch_by_proposal_len(
 
 def sampler_output_to_torch(
     sampler_output_list: List[SamplerOutput], sampler_transposed: bool
-) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor]:
+) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor, Optional[torch.Tensor]]:
     """Utility function which converts a list of SamplerOutput to tensors.
 
         sampler_transposed here is used as the indicator for whether
@@ -169,7 +169,23 @@ def sampler_output_to_torch(
     if sampler_transposed:
         sampled_token_ids = sampled_token_ids.transpose(0, 1)
 
-    return sampled_token_ids, sampled_token_probs, sampled_token_logprobs
+    if sampler_output_list[0].hidden_states is not None:
+        # shape: [batch_size, num_sampler_output, hidden_dim]
+        sampled_hidden_states = torch.stack(
+            [
+                sampler_output.hidden_states
+                for sampler_output in sampler_output_list
+            ],
+            dim=0,
+        )
+
+        if sampler_transposed:
+            sampled_hidden_states = sampled_hidden_states.transpose(0, 1)
+    else:
+        sampled_hidden_states = None
+
+    return (sampled_token_ids, sampled_token_probs, sampled_token_logprobs,
+            sampled_hidden_states)
 
 
 def maybe_mock_device_tensors(sampler_output: SamplerOutput, batch_size: int,