Dataset sampling has no shuffle / seed control
Dataset sampling has no shuffle / seed control: a task in MiMo-V2.6-RL-oss: Agentic RL Environments (MiMo RL release). I'm using lema to fine-tune on a mix of HuggingFace datasets, configured through YAML (DataParams / DatasetSplitParams / DatasetParams). I'm running into two related limitations…
The task
I'm using lema to fine-tune on a mix of HuggingFace datasets, configured through YAML (DataParams / DatasetSplitParams / DatasetParams). I'm running into two related limitations around sampling. 1. sample_count always gives me the head of the dataset When I set sample_count on…
Part of XiaomiMiMo/MiMo-V2.6-RL-oss.