xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-04-26 01:37:04 +08:00
Code Issues Packages Projects Releases Wiki Activity

[V1][Spec Decode] Update N-gram Proposer Interface (#15750)

Browse Source 
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
This commit is contained in:
Woosuk Kwon 2025-04-04 16:32:54 -07:00 committed by GitHub
parent 70ad3f9e98
commit 63375f0cdb
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 16 additions and 18 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

28
vllm/v1/spec_decode/ngram_proposer.py
View File

@ -10,14 +10,21 @@ from vllm.config import VllmConfig
class NgramProposer: class NgramProposer:
def __init__(self, vllm_config: VllmConfig): def __init__(self, vllm_config: VllmConfig):
self.vllm_config = vllm_config # Minimum length of the n-gram to match.
self.min_n = vllm_config.speculative_config.prompt_lookup_min
# Maximum length of the n-gram to match.
self.max_n = vllm_config.speculative_config.prompt_lookup_max
# Number of tokens follow the match. If there are less than k
# tokens follow the match, we will return the maximum amount of
# tokens until the end.
self.k = vllm_config.speculative_config.num_speculative_tokens
# Trigger Numba JIT compilation for N-gram proposer.
# This usually takes less than 1 second.
self.propose(np.zeros(1024, dtype=np.int32))
def propose( def propose(
self, self,
context_token_ids: np.ndarray, context_token_ids: np.ndarray,
min_n: int,
max_n: int,
k: int,
) -> Optional[np.ndarray]: ) -> Optional[np.ndarray]:
"""Proposes the next sequence of tokens based on n-gram pattern """Proposes the next sequence of tokens based on n-gram pattern
matching in the context. The function finds matches of the last n matching in the context. The function finds matches of the last n
@ -27,17 +34,12 @@ class NgramProposer:
Args: Args:
context_token_ids: Numpy array of token IDs representing the context_token_ids: Numpy array of token IDs representing the
context sequence. context sequence.
min_n: Minimum length of the n-gram to match.
max_n: Maximum length of the n-gram to match.
k: Number of tokens follow the match. If there are less
than k tokens follow the match, we will return
the maximum amount of tokens until the end.
Returns: Returns:
np.ndarray: The sequence of tokens that followed np.ndarray: The sequence of tokens that followed
the matched n-gram in the context. the matched n-gram in the context.
None: If no matching n-gram pattern is found. None: If no matching n-gram pattern is found.
Example: Example:
If context_token_ids = [1,2,3,4,2,3], min_n = 2, max_n = 3, and If context_token_ids = [1,2,3,4,2,3], min_n = 2, max_n = 3, and
k = 4: k = 4:
@ -49,8 +51,8 @@ class NgramProposer:
we only have three tokens after the match. we only have three tokens after the match.
""" """
# TODO(woosuk): Optimize this. # TODO(woosuk): Optimize this.
for n in range(max_n, min_n - 1, -1): for n in range(self.max_n, self.min_n - 1, -1):
result = _find_subarray_kmp(context_token_ids, n, k) result = _find_subarray_kmp(context_token_ids, n, self.k)
if result is not None: if result is not None:
return result return result
return None return None

6
vllm/v1/worker/gpu_model_runner.py
View File

@ -1246,11 +1246,7 @@ class GPUModelRunner(LoRAModelRunnerMixin):
end_idx = start_idx + num_sampled_ids end_idx = start_idx + num_sampled_ids
self.input_batch.token_ids_cpu[i, start_idx:end_idx] = sampled_ids self.input_batch.token_ids_cpu[i, start_idx:end_idx] = sampled_ids
drafter_output = self.drafter.propose( drafter_output = self.drafter.propose(
self.input_batch.token_ids_cpu[i, :end_idx], self.input_batch.token_ids_cpu[i, :end_idx])
self.speculative_config.prompt_lookup_min,
self.speculative_config.prompt_lookup_max,
self.speculative_config.num_speculative_tokens,
)
if drafter_output is None or len(drafter_output) == 0: if drafter_output is None or len(drafter_output) == 0:
draft_token_ids.append([]) draft_token_ids.append([])
else: else: