update topp

Signed-off-by: Yuki Huang <yukih@nvidia.com>

Author: yuki-97

tests/functional/grpo_multiturn.sh

@@ -29,7 +29,7 @@ uv run coverage run -a --data-file=$PROJECT_ROOT/tests/.coverage --source=$PROJE
     policy.max_total_sequence_length=1024 \
     policy.train_global_batch_size=4 \
     policy.train_micro_batch_size=1 \
-    policy.generation.top_p=0.99 \
+    policy.generation.top_p=0.999 \
     policy.generation.top_k=8000 \
     logger.tensorboard_enabled=true \
     logger.log_dir=$LOG_DIR \