{
  "scope": "pretrained H3 block 0 only, synthetic hidden states, no full video pipeline or distributed H3",
  "revision": "42ed227ee7df40d41602854ae760620d6eb651fe",
  "block_parameters": 645571840,
  "trainable_parameters": 100352,
  "sequence_length": 32,
  "steps": [
    {
      "step": 0,
      "loss": 1662965.375,
      "adapter_grad_norm": 121312.3203125,
      "input_grad_norm": 1392782.75,
      "seconds": 0.8394418440293521
    },
    {
      "step": 1,
      "loss": 1664710.875,
      "adapter_grad_norm": 128111.2578125,
      "input_grad_norm": 1394067.875,
      "seconds": 0.014833829016424716
    }
  ],
  "gpu_peak_allocated_bytes": 1456381440,
  "gpu": "Radeon 8060S Graphics",
  "torch_version": "2.9.1+rocm7.2.1.gitff65f5bc",
  "checkpoint_mapping": "strict keys; fused QKV split, SwiGLU gate/value swap; upstream numerical parity not yet qualified"
}