xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-13 11:35:41 +08:00
Code Issues Packages Projects Releases Wiki Activity

[V1][Spec Decoding] Include bonus tokens in mean acceptance length (#17908)

Browse Source 
Signed-off-by: Mark McLoughlin <markmc@redhat.com>
This commit is contained in:
Mark McLoughlin 2025-05-09 21:32:36 +01:00 committed by GitHub
parent ea2236bf95
commit 7e3571134f
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 9 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
examples/offline_inference/eagle.py
View File

@ -118,8 +118,8 @@ def main():
acceptance_counts[step] += count acceptance_counts[step] += count
print("-" * 50) print("-" * 50)
print(f"mean acceptance length: \ print(f"mean acceptance length (including bonus tokens): \
{sum(acceptance_counts) / acceptance_counts[0]:.2f}") {1 + (sum(acceptance_counts) / acceptance_counts[0]):.2f}")
print("-" * 50) print("-" * 50)
# print acceptance at each token position # print acceptance at each token position

10
vllm/v1/spec_decode/metrics.py
View File

@ -73,7 +73,9 @@ class SpecDecodingLogging:
draft_acceptance_rate = (num_accepted_tokens / num_draft_tokens * draft_acceptance_rate = (num_accepted_tokens / num_draft_tokens *
100 if num_draft_tokens > 0 else float("nan")) 100 if num_draft_tokens > 0 else float("nan"))
mean_acceptance_length = (num_accepted_tokens / num_drafts)
# Conventionally, mean acceptance length includes the bonus token
mean_acceptance_length = 1 + (num_accepted_tokens / num_drafts)
pos_matrix = np.array(self.accepted_tokens_per_pos_lists) pos_matrix = np.array(self.accepted_tokens_per_pos_lists)
acceptance_rates = np.sum(pos_matrix, axis=0) / num_drafts acceptance_rates = np.sum(pos_matrix, axis=0) / num_drafts
@ -103,10 +105,12 @@ class SpecDecodingProm:
rate(vllm:spec_decode_num_accepted_tokens_total[$interval]) / rate(vllm:spec_decode_num_accepted_tokens_total[$interval]) /
rate(vllm:spec_decode_num_draft_tokens_total[$interval]) rate(vllm:spec_decode_num_draft_tokens_total[$interval])
The mean acceptance length can be calculated using: The mean acceptance length (conventionally including bonus tokens)
can be calculated using:
1 + (
rate(vllm:spec_decode_num_accepted_tokens_total[$interval]) / rate(vllm:spec_decode_num_accepted_tokens_total[$interval]) /
rate(vllm:spec_decode_num_drafts[$interval]) rate(vllm:spec_decode_num_drafts[$interval]))
A per-position acceptance rate vector can be computed using A per-position acceptance rate vector can be computed using