agent-lightning
agent-lightning copied to clipboard
microsoft
The absolute trainer to light up AI agents.
## Summary - allow `training_rollout_batch`/`validation_rollout_batch` APIs to receive per-task resource lists - add `batch_size` to `AgentRunner` and propagate through `Trainer` for batched polling - test batch rollouts with resource lists...
## Summary - allow passing defaults for lightning_cli arguments - support programmatic defaults overriding required CLI args - test default overrides for required parameters ## Testing - `pytest` ------ https://chatgpt.com/codex/tasks/task_e_689590e72300832e95482ca7a6b7e4f8
The credit assignment in the sample code applies a uniform policy to all states. However, I'd like to assign rewards differently based on the specific action taken. What would be...
I see the current implementation is written in Python, but I’m wondering about framework support. Are there any plans to support TypeScript/JavaScript-based frameworks as well? Or would it be possible...
We can't get token ids from LLM Proxy when stream is enabled. vLLM has token_ids returned in streaming. The problem is with LiteLLM and has three parts. 1. `.venv/lib/python3.12/site-packages/litellm/litellm_core_utils/streaming_handler.py` has...
## Summary This PR refactors the `LLMProxy.start()` logic to launch the Uvicorn proxy server in a fully isolated process using `multiprocessing.spawn`. The previous implementation ran the server in a background...
The linked Discord Chat seems to be wrong. The link leads to a Discord chat with a different topic.
Like does it fine tune the model or does it adjust the system prompt or do anything else . like how does the agent learn becaue agent can get intelligence...
