{
  "campaign": "runs/campaign",
  "identity": "fb2763cee04396e032383cdcd081266340ce7e7e69f5a70da87b5512efeac4af",
  "source_revision": "4f6e2f4061f909ff95aedaffc7e336d06011c603",
  "source_hash": "f14fc16e8fc3d2e947f00d80af50a6dafe5d4e926cd48734f728c6d1199c9ce4",
  "cache_identity": "8cf0c4883c19d26f378623f91bfe079bc8ebf8f23e51814821f4c8c28c4b8151",
  "seed": 42,
  "loader_version": 1,
  "validation_tokens": 197411295,
  "validation_complete": true,
  "epochs_per_run": 40,
  "selection": "final checkpoint full validation cross-entropy",
  "runs": [
    {
      "run": "20m-lr0.001-wd0.1",
      "preset": "20m",
      "parameters": 20403520,
      "lr": 0.001,
      "weight_decay": 0.1,
      "beta2": 0.95,
      "loss": 3.6097478024886414,
      "tokens": 408071168,
      "peak_memory_bytes": 42391254016,
      "median_tokens_per_second": 1528594.9343495017,
      "seed": 42,
      "deterministic": false,
      "checkpoint": "runs/campaign/20m-lr0.001-wd0.1/final.pt"
    },
    {
      "run": "20m-lr0.003-wd0.1",
      "preset": "20m",
      "parameters": 20403520,
      "lr": 0.003,
      "weight_decay": 0.1,
      "beta2": 0.95,
      "loss": 3.6135744227963267,
      "tokens": 408071168,
      "peak_memory_bytes": 42391254016,
      "median_tokens_per_second": 1513908.3178001884,
      "seed": 42,
      "deterministic": false,
      "checkpoint": "runs/campaign/20m-lr0.003-wd0.1/final.pt"
    },
    {
      "run": "20m-lr0.001-wd0",
      "preset": "20m",
      "parameters": 20403520,
      "lr": 0.001,
      "weight_decay": 0.0,
      "beta2": 0.95,
      "loss": 3.621142972280689,
      "tokens": 408071168,
      "peak_memory_bytes": 42391254016,
      "median_tokens_per_second": 1515886.6160321503,
      "seed": 42,
      "deterministic": false,
      "checkpoint": "runs/campaign/20m-lr0.001-wd0/final.pt"
    },
    {
      "run": "20m-lr0.003-wd0",
      "preset": "20m",
      "parameters": 20403520,
      "lr": 0.003,
      "weight_decay": 0.0,
      "beta2": 0.95,
      "loss": 3.622035902191771,
      "tokens": 408071168,
      "peak_memory_bytes": 42391254016,
      "median_tokens_per_second": 1522494.9311930325,
      "seed": 42,
      "deterministic": false,
      "checkpoint": "runs/campaign/20m-lr0.003-wd0/final.pt"
    },
    {
      "run": "20m-lr0.0003-wd0.1",
      "preset": "20m",
      "parameters": 20403520,
      "lr": 0.0003,
      "weight_decay": 0.1,
      "beta2": 0.95,
      "loss": 3.742181285807369,
      "tokens": 408071168,
      "peak_memory_bytes": 42391483392,
      "median_tokens_per_second": 1528131.629942356,
      "seed": 42,
      "deterministic": false,
      "checkpoint": "runs/campaign/20m-lr0.0003-wd0.1/final.pt"
    },
    {
      "run": "20m-lr0.0003-wd0",
      "preset": "20m",
      "parameters": 20403520,
      "lr": 0.0003,
      "weight_decay": 0.0,
      "beta2": 0.95,
      "loss": 3.7497837995628553,
      "tokens": 408071168,
      "peak_memory_bytes": 42391483392,
      "median_tokens_per_second": 1520907.4931353792,
      "seed": 42,
      "deterministic": false,
      "checkpoint": "runs/campaign/20m-lr0.0003-wd0/final.pt"
    },
    {
      "run": "50m-parent0-beta20.99",
      "preset": "50m",
      "parameters": 48507392,
      "lr": 0.001,
      "weight_decay": 0.1,
      "beta2": 0.99,
      "loss": 3.257810754200342,
      "tokens": 970147840,
      "peak_memory_bytes": 42918305792,
      "median_tokens_per_second": 843483.712657399,
      "seed": 42,
      "deterministic": false,
      "checkpoint": "runs/campaign/50m-parent0-beta20.99/final.pt"
    },
    {
      "run": "50m-parent1-beta20.99",
      "preset": "50m",
      "parameters": 48507392,
      "lr": 0.003,
      "weight_decay": 0.1,
      "beta2": 0.99,
      "loss": 3.2911234982623014,
      "tokens": 970147840,
      "peak_memory_bytes": 42918305792,
      "median_tokens_per_second": 843098.2392579638,
      "seed": 42,
      "deterministic": false,
      "checkpoint": "runs/campaign/50m-parent1-beta20.99/final.pt"
    },
    {
      "run": "50m-parent0-beta20.95",
      "preset": "50m",
      "parameters": 48507392,
      "lr": 0.001,
      "weight_decay": 0.1,
      "beta2": 0.95,
      "loss": 3.293054120906817,
      "tokens": 970147840,
      "peak_memory_bytes": 42918305792,
      "median_tokens_per_second": 842963.8874621252,
      "seed": 42,
      "deterministic": false,
      "checkpoint": "runs/campaign/50m-parent0-beta20.95/final.pt"
    },
    {
      "run": "50m-parent1-beta20.95",
      "preset": "50m",
      "parameters": 48507392,
      "lr": 0.003,
      "weight_decay": 0.1,
      "beta2": 0.95,
      "loss": 3.328141890581081,
      "tokens": 970147840,
      "peak_memory_bytes": 42918305792,
      "median_tokens_per_second": 837629.0473204984,
      "seed": 42,
      "deterministic": false,
      "checkpoint": "runs/campaign/50m-parent1-beta20.95/final.pt"
    },
    {
      "run": "90m-parent0",
      "preset": "90m",
      "parameters": 91605120,
      "lr": 0.001,
      "weight_decay": 0.1,
      "beta2": 0.99,
      "loss": 3.0493794146794455,
      "tokens": 1832102912,
      "peak_memory_bytes": 43707437056,
      "median_tokens_per_second": 496751.1402128865,
      "seed": 42,
      "deterministic": false,
      "checkpoint": "runs/campaign/90m-parent0/final.pt"
    },
    {
      "run": "90m-parent1",
      "preset": "90m",
      "parameters": 91605120,
      "lr": 0.003,
      "weight_decay": 0.1,
      "beta2": 0.99,
      "loss": 3.122935172801696,
      "tokens": 1832102912,
      "peak_memory_bytes": 43707437056,
      "median_tokens_per_second": 492034.8140604545,
      "seed": 42,
      "deterministic": false,
      "checkpoint": "runs/campaign/90m-parent1/final.pt"
    }
  ],
  "limitation": "Single seed; 90M compared learning rates with beta2=0.99 only, so improvement over beta2=0.95 was not measured."
}
