Get Zlio on your home screenOne tap, full screen, like an app.
Developer Tools · first seen 13 November 2026, 21:09 UTC
Hugging Face fixes TRL bug reusing stale KV cache
TRL v1.18.0 adds a model-version cache salt so AsyncGRPO and AsyncDistillation student rollouts stop pairing new weights with first-version prefix KV cache from older weights.