research/problems/cross_entropy/resources/submission_spec.json

{
  "problem_name": "cross_entropy",
  "description": "Triton kernel optimization problem for high-performance cross entropy loss computation",
  "requirements": {
    "cuda_backend": true,
    "gpu_required": true,
    "triton_version": ">=2.1.0",
    "torch_version": ">=2.0.0"
  },
  "evaluation": {
    "timeout_seconds": 300,
    "memory_limit_mb": 8192,
    "gpu_memory_limit_mb": 4096
  },
  "metadata": {
    "M_list": [256, 512, 1024],
    "N": 8192
  }
}