Skip to content
Merged
Changes from all commits
Commits
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
Original file line number Diff line number Diff line change
Expand Up @@ -17,7 +17,6 @@
from typing import Any, Optional

import ray
import torch
from omegaconf import DictConfig

from verl.experimental.agent_loop.agent_loop import (
Expand Down Expand Up @@ -96,11 +95,10 @@ async def generate(
final_output.token_ids.extend(output.token_ids)
if output.log_probs is not None:
final_output.log_probs.extend(output.log_probs)
# sglang returns routed_experts for the full sequence (prompt + all tokens),
# so on partial rollout resume the new output already covers all positions.
if output.routed_experts is not None:
if final_output.routed_experts is None:
final_output.routed_experts = output.routed_experts
else:
final_output.routed_experts = torch.cat([final_output.routed_experts, output.routed_experts], dim=0)
Comment thread
NoonePauseferg marked this conversation as resolved.
final_output.routed_experts = output.routed_experts
if output.num_preempted is not None:
final_output.num_preempted += output.num_preempted
final_output.stop_reason = output.stop_reason
Expand Down
Loading