agent-lightning
agent-lightning copied to clipboard
microsoft
The absolute trainer to light up AI agents.
## Summary - add a contributor guide detailing environment setup, branching conventions, and PR expectations - document the maintainer release workflow, CI label usage, and backport process - surface both...
## Summary - capture the raw `uv lock --upgrade` output for use in the automated PR description - timestamp the generated branch and PR title while updating the schedule to...
Issue: agentops mock service and OTL batchprocessor shutdown are not synchronized: 🖇 AgentOps: Network error during span export: HTTPConnectionPool(host='localhost', port=41733): Max retries exceeded with url: /traces (Caused by NewConnectionError(': Failed...
The Error traceback: ``` File "/home/tiger/.pyenv/versions/3.11.2/lib/python3.11/site-packages/agentlightning/verl/entrypoint.py", line 152, in run trainer.fit() File "/home/tiger/.pyenv/versions/3.11.2/lib/python3.11/site-packages/agentlightning/verl/trainer.py", line 318, in fit metrics = self._train_step(batch_dict) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/home/tiger/.pyenv/versions/3.11.2/lib/python3.11/site-packages/agentlightning/verl/trainer.py", line 95, in _train_step batch, agent_metrics =...
1. The model file produced after training is in FSDP format (designed for large-scale distributed training and saved by PyTorch FSDP after training), which is different from HuggingFace (a universal...
I can see the example of search_r1 is for v1 api, such as n_workers (n_runners in v2) and url training dataset? Can you explain the new features? Can you provide...
Great work! I would like to know whether this framework supports multimodal input from agents. For example, could it handle image and text responses from agents (perhaps similar to OpenAI...
## Reproduce bug ### How to insert `prompt = " ".join([prompt for _ in range(3000)])` below [this line](https://github.com/microsoft/agent-lightning/blob/d00cc3e4aab227207520c080e30003c5597a988c/examples/calc_x/calc_agent.py#L114). ### Full trace stack on server side ```base ray.exceptions.RayTaskError(RuntimeError): ray::TaskRunner.run() (pid=2135564, ip=10.8.163.2,...
Modelo Modular de Conciencia y Meta-Conciencia Funcional --- Definición matemática del modelo Y = \begin{cases} f_{modular}^{(n)} = \sum_{i \in activos} w_i M_i^{(n)}, \ M_X^{(n+1)} = 6 - M_X^{(n)}, \quad...
