Skip to content

Fix prefill_inference #13885

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 1 commit into from
Sep 2, 2025
Merged

Fix prefill_inference #13885

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
9 changes: 3 additions & 6 deletions examples/qualcomm/oss_scripts/llama/decoder_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -83,7 +83,6 @@ def _model_call(self, inps):
inps,
self._model,
self._tokenizer,
self.ar_len,
self.max_seq_length,
use_i64_token=self.use_i64_token,
collect_logits=True,
Expand Down Expand Up @@ -458,15 +457,13 @@ def prefill_inference(
logits, new_k_caches, new_v_caches = results
elif len(results) == 1:
logits = results
logits = torch.argmax(logits[:, pos - 1], dim=-1).item()
token_list.append(logits)
token = torch.argmax(logits[:, pos - 1], dim=-1).item()
token_list.append(token)
if collect_logits:
result_logits.append(logits)
result_logits = logits[:, :pos]
pos += 1

logging.info(f"prefill inference result:\n{tokenizer.decode(token_list)}")
if collect_logits:
result_logits = torch.cat(result_logits, dim=1)
return result_logits


Expand Down
Loading