Skip to content

[fix] Add 1 and draft_token_num to seq_len when overlap scheduling is enabled during memory estimation #5343

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 6 commits into from
Jun 24, 2025
Merged

[fix] Add 1 and draft_token_num to seq_len when overlap scheduling is enabled during memory estimation #5343

Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
00702a1
Add one to seq_len for overlap during memory estimation
HuiGao-NV Jun 19, 2025
bb528cf
Remove file added by mistake
HuiGao-NV Jun 19, 2025
d7cf3d9
Add max_draft_tokens more tokens to kv cache seq_len for memory
HuiGao-NV Jun 19, 2025
0252a9b
Merge branch 'main' into extra_token_for_overlap
HuiGao-NV Jun 20, 2025
3a19409
Merge branch 'main' into extra_token_for_overlap
HuiGao-NV Jun 23, 2025
863999a
Merge branch 'main' into extra_token_for_overlap
HuiGao-NV Jun 24, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 6 additions & 0 deletions tensorrt_llm/_torch/pyexecutor/_util.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -151,7 +151,13 @@ def _get_token_num_for_estimation(self) -> int:
# estimate_max_kv_cache_tokens submits self._dummy_reqs
num_cache_blocks = 0
num_extra_tokens_per_seq = 1 # account for generated tokens
pytorch_backend_config = executor_config.pytorch_backend_config
spec_cfg = executor_config.speculative_config
if not pytorch_backend_config.disable_overlap_scheduler:
num_extra_tokens_per_seq = num_extra_tokens_per_seq + 1
if spec_cfg is not None:
num_extra_tokens_per_seq += spec_cfg.max_draft_tokens

if spec_cfg is not None:
num_extra_tokens_per_seq += spec_cfg.max_draft_tokens
num_extra_tokens_per_seq += spec_cfg.num_extra_kv_tokens
Expand Down