agent-lightning · Issues· 160 open
Open on GitHubLocally synced open issues (discussions stay on GitHub)
- #593
per_rollout_mean divides by training rows instead of rollouts, so the loss scale changes when a rollout spans several rows
Updated Sep 17, 2026 - #236
Agent Lightning WeChat Discuss Group
Updated Sep 14, 2026 - #592
verl 0.7.x embedded vLLM server cannot load custom tool parser plugins (tool_parser_plugin import step skipped)
Updated Sep 14, 2026 - #471
Encountered a problem "Unclosed client session"
bugexamplestopic/ragv0.xUpdated Sep 8, 2026 - #566
Design: ThinkingBox x Agent Lightning stateful MCP training example
Updated Sep 4, 2026 - #69
How to evaluate the trained model?
help wantedquestionverlv0.xUpdated Sep 4, 2026 - #567
Add CPU-side Kubernetes image-readiness preflight for VERL training
Updated Aug 29, 2026 - #570
Documentation menu hides under warning banner
Updated Aug 27, 2026 - #504
RAG training becomes slow over time with frequent MCP retrieve timeouts
v0.xUpdated Aug 25, 2026 - #490
Exploration collapse: agents converge to fixed policies, losing novel behavior discovery
enhancementv0.xUpdated Aug 25, 2026 - #464
Which hyper-parameters in the standard VERL configuration do not take effect?
documentationenhancementhelp wantedv0.xUpdated Aug 25, 2026 - #456
about llm endpoint
questionverlproxyv0.xUpdated Aug 25, 2026 - #448
How to deal with failed rollouts?
documentationhelp wantedquestionv0.xUpdated Aug 25, 2026 - #439
Would it be possible to allow custom text gradient computation and editing templates for APO in the form of SDK usage?
enhancementapov0.xUpdated Aug 25, 2026 - #436
Regarding the server's lifecycle
questionstorev0.xUpdated Aug 25, 2026