HF RL Explorer

I want both the TaskPrompter and InvPT evaluation packages to expose a stateful…

I want both the TaskPrompter and InvPT evaluation packages to expose a stateful…: a task in MiMo-V2.6-RL-oss: Agentic RL Environments (MiMo RL release). A caller should create a meter, call update(preds, target) any number of times with PyTorch tensors shaped like saliency maps, and then call get…

The task

A caller should create a meter, call update(preds, target) any number of times with PyTorch tensors shaped like saliency maps, and then call get_score(verbose=False) to receive a dictionary containing maxF. Predictions are image-scale saliency values: update should convert them…

Part of XiaomiMiMo/MiMo-V2.6-RL-oss.