mamba-d1-4j2vbaoi/meta_002500.json

{
  "step": 2500,
  "val_bpb": 2.3049861890100987,
  "model_config": {
    "sequence_len": 2048,
    "vocab_size": 265,
    "n_layer": 1,
    "n_head": 16,
    "n_kv_head": 16,
    "n_embd": 1536,
    "adaptive_decay": false,
    "rope": true,
    "repetition_count": 1,
    "mamba_layers": [
      0
    ],
    "recurrent_vocab_layers": [],
    "attention_time_decay": false
  },
  "user_config": {
    "run": "mamba-d1",
    "device_type": "cuda",
    "depth": 1,
    "repetition_count": 1,
    "attention_time_decay": false,
    "adaptive_decay": false,
    "rope": true,
    "init_from_tag": "",
    "resume_from_tag": "",
    "resume_from_step": -1,
    "resume_from_hub": true,
    "resume_hub_repo_id": "ASzecsenyi/VQLM",
    "resume_hub_subdir": "",
    "resume_hub_repo_type": "model",
    "num_iterations": -1,
    "target_flops": -1.0,
    "target_param_data_ratio": 20,
    "num_epochs": -1,
    "data": "base_data",
    "max_seq_len": 2048,
    "device_batch_size": 4,
    "total_batch_size": -1,
    "embedding_lr": 0.2,
    "unembedding_lr": 0.004,
    "weight_decay": 0.0,
    "matrix_lr": 0.02,
    "base_lr": 1.0,
    "grad_clip": 1.0,
    "warmup_ratio": 0.05,
    "warmdown_ratio": 0.2,
    "final_lr_frac": 0.0,
    "eval_every": 2500,
    "eval_tokens": 10485760,
    "core_metric_every": 2000,
    "core_metric_max_per_task": 500,
    "sample_every": 2500,
    "checkpoint_every": -1,
    "max_checkpoints": 3,
    "push_checkpoints_to_hub": true,
    "hf_repo_id": "ASzecsenyi/VQLM",
    "hf_repo_type": "model",
    "model_tag": "mamba-d1",
    "ddp": true,
    "ddp_rank": 0,
    "ddp_local_rank": 0,
    "ddp_world_size": 1
  },
  "device_batch_size": 4,
  "max_seq_len": 2048,
  "num_iterations": 36885,
  "warmdown_ratio": 0.2,
  "max_checkpoints": 3,
  "push_checkpoints_to_hub": true,
  "hf_repo_id": "ASzecsenyi/VQLM",
  "hf_repo_type": "model"
}