benchmark/denoiser/shared-activation-runtime.json

{
  "deterministic": true,
  "gpu": "NVIDIA A40",
  "median_seconds": 47.88486023619771,
  "model": "WaveCut/Krea-2-Turbo-OrbitQuant-W4A4",
  "output_hash": "7eebffa5bc2438c64f2f7fed397591d80dee6e8c665358ac19cc93fb0cf4c87e",
  "peak_allocated_mib": 12895.21728515625,
  "peak_reserved_mib": 14978.0,
  "protection": null,
  "restore_img_in": false,
  "revision": "ba2135909967b8038009d98d710ceaf1e137ff8e",
  "runtime_inventory": {
    "activation_kernel_backends": [
      "triton_cuda",
      "triton_cuda_int8_surrogate",
      "triton_cuda_packed_w4"
    ],
    "class_name": "Krea2Transformer2DModel",
    "dequantized_cache_bytes": 0,
    "effective_runtime_modes": [
      "native_packed_matmul"
    ],
    "executed_orbitquant_linear_count": 233,
    "int8_weight_cache_bytes": 0,
    "last_activation_cache_hits": 120,
    "orbitquant_linear_count": 233,
    "source_precision_linear_count": 31
  },
  "samples_seconds": [
    47.73804524168372,
    47.88486023619771,
    47.920409340411425
  ],
  "shape": {
    "height": 2048,
    "steps": 8,
    "width": 2048
  },
  "status": "pass",
  "torch": "2.8.0+cu128",
  "warm_output_hashes": [
    "7eebffa5bc2438c64f2f7fed397591d80dee6e8c665358ac19cc93fb0cf4c87e"
  ]
}