NeoLLM / optimizer_precision.json
KitsuVp's picture
Model save
d6b5dd0 verified
Raw
History Blame Contribute Delete
864 Bytes
{
"ademamix_enabled": true,
"bf16_stochastic_round_active": false,
"bf16_stochastic_round_elements": 0,
"bf16_stochastic_round_groups": 0,
"bf16_stochastic_round_tensors": 0,
"delta_eligible_linear_weights": 0,
"delta_enabled": false,
"delta_eta": 0.37,
"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": false,
"mxfp8_available": false,
"mxfp8_linear_layers": 0,
"mxfp8_recipe": null,
"native_precision_elements": 0,
"optimizer_variant": "AdEMAMix",
"ordinary_linear_layers": 124,
"parameter_elements_by_dtype": {
"bfloat16": 84565816,
"float32": 2112
},
"parameter_storage_bytes": 169140080,
"runtime_precision_mode": "BF16_FP32_MASTER",
"stochastic_round_seed": 0
}