HF RL Explorer

Repair ONNX TensorScatter circular KV-cache indexing and run the offline reference inference fixture.

Repair ONNX TensorScatter circular KV-cache indexing and run the offline reference inference fixture.: a task in MiMo-V2.6-RL-oss: Agentic RL Environments (MiMo RL release). The workspace under /app/vendor/onnx is a frozen slice of a tensor operator used for bounded attention KV-cache updates…

The task

The workspace under /app/vendor/onnx is a frozen slice of a tensor operator used for bounded attention KV-cache updates. Diagnose and repair the circular-mode indexing bug in the existing production files: vendor/onnx/onnx/reference/ops/op_tensor_scatter.py…

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