base_checkpoints/d34/meta_010507.json

{
  "step": 10507,
  "val_bpb": 0.7259969166510065,
  "model_config": {
    "sequence_len": 2048,
    "vocab_size": 32768,
    "n_layer": 34,
    "n_head": 17,
    "n_kv_head": 17,
    "n_embd": 2176,
    "window_pattern": "L"
  },
  "user_config": {
    "run": "pre1900_d34_full22btok",
    "device_type": "",
    "fp8": true,
    "fp8_recipe": "tensorwise",
    "depth": 34,
    "aspect_ratio": 64,
    "head_dim": 128,
    "max_seq_len": 2048,
    "window_pattern": "L",
    "num_iterations": -1,
    "target_flops": -1.0,
    "target_param_data_ratio": 11.0,
    "device_batch_size": 4,
    "total_batch_size": -1,
    "embedding_lr": 0.3,
    "unembedding_lr": 0.004,
    "weight_decay": 0.2,
    "matrix_lr": 0.02,
    "scalar_lr": 0.5,
    "adam_beta1": 0.8,
    "adam_beta2": 0.95,
    "warmup_ratio": 0.0,
    "warmdown_ratio": 0.5,
    "final_lr_frac": 0.0,
    "resume_from_step": -1,
    "eval_every": 250,
    "eval_tokens": 20971520,
    "core_metric_every": 2000,
    "core_metric_max_per_task": 500,
    "sample_every": 2000,
    "save_every": 3000,
    "model_tag": null
  },
  "device_batch_size": 4,
  "max_seq_len": 2048,
  "dataloader_state_dict": {
    "pq_idx": 201,
    "rg_idx": 0,
    "epoch": 1
  },
  "loop_state": {
    "min_val_bpb": 0.7259969166510065,
    "smooth_train_loss": 2.3119625572178344,
    "total_training_time": 51539.05968785286
  }
}