Skip to content

Add top probabilities to the output #2365

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

Open
Dakantz wants to merge 2 commits into
base: main
Choose a base branch
from
+8 −0
Open

Add top probabilities to the output #2365

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
b285176
Add all log probabilites directly!
Dakantz May 2, 2026
3f800a7
Merge branch 'abetlen:main' into main
Dakantz Sep 5, 2026
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
8 changes: 8 additions & 0 deletions llama_cpp/llama.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1733,6 +1733,7 @@ def logit_bias_processor(
token_logprobs: List[Optional[float]] = []
tokens: List[str] = []
top_logprobs: List[Optional[Dict[str, float]]] = []
top_logprob_tokens: List[Optional[Dict[int, float]]] = []

if echo:
# Remove leading BOS token if exists
Expand Down Expand Up @@ -1777,8 +1778,13 @@ def logit_bias_processor(
): logprob
for logprob, i in sorted_logprobs[:logprobs]
}
top_logprob_token: Optional[Dict[int, float]] = {
i: logprob for logprob, i in sorted_logprobs[:logprobs]
}
top_logprob.update({token_str: logprobs_token[int(token)]})
top_logprob_token.update({token: logprobs_token[int(token)]})
top_logprobs.append(top_logprob)
top_logprob_tokens.append(top_logprob_token)
# Weird idosincracy of the OpenAI API where
# token_logprobs and top_logprobs are null for
# the first token.
Expand All @@ -1790,6 +1796,8 @@ def logit_bias_processor(
"text_offset": text_offsets,
"token_logprobs": token_logprobs,
"top_logprobs": top_logprobs,
"top_logprob_tokens": top_logprob_tokens,
"all_logprobs": all_logprobs,
}

yield {
Expand Down