{
 "buffer_lines": [
  "load_tensors:   CPU_Mapped model buffer size = 35835.49 MiB",
  "load_tensors:        CUDA0 model buffer size = 18870.50 MiB",
  "load_tensors:   CPU_Mapped model buffer size = 27465.95 MiB"
 ],
 "cuda_mib": [
  18870.5
 ],
 "device_lines": [
  "using device CUDA0 (NVIDIA GeForce RTX 3090 Ti)"
 ],
 "expect": {
  "n_cpu_moe": 34
 },
 "host_mib": [
  35835.49,
  27465.95
 ],
 "lazy_lines": [
  "add: tensor per_layer_token_embd.weight (size = 27465 MiB) lazy read enabled"
 ],
 "offload_lines": [
  "offloaded 49/49 layers to GPU"
 ],
 "verdict": "hybrid (--n-cpu-moe 34)",
 "why": "the registered line keeps 34 expert layers in host memory: recorded, never judged whole; offloaded 49/49; buffers ['load_tensors:   CPU_Mapped model buffer size = 35835.49 MiB', 'load_tensors:        CUDA0 model buffer size = 18870.50 MiB', 'load_tensors:   CPU_Mapped model buffer size = 27465.95 MiB']"
}