Skip to content

Fix: Prevent infinite “A5A5A5...” repetition loop in generate() (Issue #1008) #1020

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
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Fix: Prevent infinite “A5A5A5...” repetition loop in generate() (Issue #1008) #1020

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
19 changes: 19 additions & 0 deletions inference/generate.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,6 +9,7 @@
from safetensors.torch import load_model

from model import Transformer, ModelArgs
import re


def sample(logits, temperature: float = 1.0):
Expand Down Expand Up @@ -67,6 +68,24 @@ def generate(
tokens[:, cur_pos] = next_token
finished |= torch.logical_and(~prompt_mask[:, cur_pos], next_token == eos_id)
prev_pos = cur_pos
# === FIX for DeepSeek-V3 Issue #1008 ===
# Detect infinite repeating loops like "A5A5A5A5..." or same token repetition

# Convert current generated tokens to string for pattern check
decoded_output = "".join([str(t) for t in tokens[0, :cur_pos].tolist()])

# (a) Detect if "A5" repeats 10 or more times continuously
if re.search(r"(A5){10,}", decoded_output):
print("[Warning] Detected excessive 'A5' repetition — stopping generation early.")
break

# (b) Detect same token repeated multiple times (generic loop check)
if cur_pos > 10:
last_10 = tokens[0, cur_pos-10:cur_pos].tolist()
if len(set(last_10)) == 1:
print("[Warning] Detected same token repetition — stopping generation early.")
break
# === END FIX ===
if finished.all():
break
completion_tokens = []
Expand Down