Skip to content
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

[BugFix] Fix cuda graph for MLPSpeculator #5875

Merged
merged 1 commit into from
Jun 27, 2024
Merged

[BugFix] Fix cuda graph for MLPSpeculator #5875

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
1 change: 0 additions & 1 deletion examples/offline_inference_mlpspeculator.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -52,7 +52,6 @@ def time_generation(llm: LLM, prompts: List[str],
speculative_model="ibm-fms/llama-13b-accelerator",
# These are currently required for MLPSpeculator decoding
use_v2_block_manager=True,
enforce_eager=True,
)

print("With speculation")
Expand Down
9 changes: 6 additions & 3 deletions vllm/worker/model_runner.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1020,10 +1020,13 @@ def execute_model(

if self.return_hidden_states:
# we only need to pass hidden states of most recent token
assert model_input.sampling_metadata is not None
indices = model_input.sampling_metadata.selected_token_indices
if model_input.is_prompt:
assert model_input.sampling_metadata is not None
hidden_states = hidden_states.index_select(
0, model_input.sampling_metadata.selected_token_indices)
hidden_states = hidden_states.index_select(0, indices)
elif decode_meta.use_cuda_graph:
hidden_states = hidden_states[:len(indices)]

output.hidden_states = hidden_states

return output
Expand Down
Loading