Skip to content

[None][feat] Eagle, use last hidden post norm #7546

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 2 commits into from
Sep 15, 2025
Merged

[None][feat] Eagle, use last hidden post norm #7546

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
e671c8d
Eagle, use last hidden post norm
IzzyPutterman Sep 4, 2025
03697e9
small fix
IzzyPutterman Sep 9, 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
5 changes: 5 additions & 0 deletions tensorrt_llm/_torch/models/modeling_speculative.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -410,6 +410,7 @@ def __init__(self, model: TModel, model_config: ModelConfig[TConfig]):
assert key in ('attn_layers', 'mla_layers')
assert key in model_config.extra_attrs
model_config.extra_attrs[key].update(value)
self.layer_idx = -1

def forward(
self,
Expand All @@ -430,6 +431,10 @@ def forward(
**kwargs,
)

if spec_metadata is not None and spec_metadata.is_layer_capture(
self.layer_idx):
spec_metadata.maybe_capture_hidden_states(self.layer_idx,
hidden_states)
if attn_metadata.padded_num_tokens is not None:
hidden_states = hidden_states[:attn_metadata.num_tokens]

Expand Down
6 changes: 4 additions & 2 deletions tensorrt_llm/_torch/speculative/eagle3.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -94,8 +94,10 @@ class Eagle3SpecMetadata(SpecMetadata):
eagle3_resource_manager: Optional[Eagle3ResourceManager] = None

def __post_init__(self):
if self.layers_to_capture is None:
if self.is_draft_model or self.num_layers == 1:
if self.is_draft_model:
self.layers_to_capture = (self.num_layers - 1, )
elif self.layers_to_capture is None:
if self.num_layers == 1:
self.layers_to_capture = (self.num_layers - 1, )
else:
if self.num_layers <= 5:
Expand Down
Loading