Skip to content

[DSV4.1][gfx1250] Fix the crash on a single-launch served step: no hidden states for the drafter's ids (#3177) #3185

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
valarLip merged 1 commit into from
Nov 11, 2026
Merged

[DSV4.1][gfx1250] Fix the crash on a single-launch served step: no hidden states for the drafter's ids (#3177) #3185

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
23 changes: 14 additions & 9 deletions atom/model_engine/model_runner.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3787,9 +3787,7 @@ def postprocess(
self.tokenID_processor.prev_token_ids = next_token_ids
draft_token_ids = self.propose_draft_token_ids(
batch,
self.tokenID_processor.input_ids.gpu[
1 : hidden_states.shape[0] + 1
],
self._draft_input_ids(batch, hidden_states),
hidden_states,
next_token_ids,
num_reject_tokens,
Expand Down Expand Up @@ -3820,9 +3818,7 @@ def postprocess(
# propose_draft_token_ids, from `batch.next_token_ids`.
draft_token_ids = self.propose_draft_token_ids(
batch,
self.tokenID_processor.input_ids.gpu[
1 : hidden_states.shape[0] + 1
],
self._draft_input_ids(batch, hidden_states),
hidden_states,
next_token_ids,
num_reject_tokens,
Expand Down Expand Up @@ -3961,9 +3957,7 @@ def forward(self, batch: ScheduledBatch) -> ScheduledBatchOutput:
if will_align_draft:
self.propose_draft_token_ids(
batch,
self.tokenID_processor.input_ids.gpu[
1 : hidden_states.shape[0] + 1
],
self._draft_input_ids(batch, hidden_states),
hidden_states,
torch.zeros(
batch.total_seqs_num, dtype=torch.int32, device=self.device
Expand Down Expand Up @@ -4072,6 +4066,17 @@ def async_proc_aggregation(self) -> KVConnectorOutput:
finished_sending=done_sending, finished_recving=done_recving
)

def _draft_input_ids(self, batch: ScheduledBatch, hidden_states) -> torch.Tensor:
"""The target ids the drafter reads, one per row the forward ran: its
`hidden_states`' rows (a PCP rank's own shard on a pure middle chunk),
else the step's `running_tokens` -- a step the model served whole
(`step_override`) returns no hidden states and drafts by itself."""
if hidden_states is None:
rows = get_forward_context().context.forward_mode.running_tokens
else:
rows = hidden_states.shape[0]
return self.tokenID_processor.input_ids.gpu[1 : rows + 1]

def propose_draft_token_ids(
self,
batch: ScheduledBatch,
Expand Down
Loading