ademamix_precision.json

{
  "bf16_stochastic_round_active": false,
  "bf16_stochastic_round_elements": 0,
  "bf16_stochastic_round_groups": 0,
  "bf16_stochastic_round_tensors": 0,
  "fp32_master_active": true,
  "fp32_master_elements": 84567928,
  "fp32_master_groups": 6,
  "fp32_master_tensors": 366,
  "fp32_passthrough_elements": 0,
  "mode": "fp32_master",
  "mxfp8_active": true,
  "mxfp8_available": true,
  "mxfp8_linear_layers": 72,
  "mxfp8_recipe": "MXFP8_RCEIL",
  "native_precision_elements": 0,
  "ordinary_linear_layers": 52,
  "parameter_elements_by_dtype": {
    "bfloat16": 84565816,
    "float32": 2112
  },
  "parameter_storage_bytes": 169140080,
  "runtime_precision_mode": "MXFP8_RCEIL_FP32_MASTER",
  "stochastic_round_seed": 0
}