understanding_good_stuff.json

{
  "num_heads": 16,
  "attention_dim": 1024,
  "vocab_size": 50260,
  "num_blocks": 24,
  "ff_dim": 3072,
  "dropout_rate": 0.1,
  "possible_opt_path": "Loaded_model",
  "max_len": 2048,
  "emb_splt": 256,
  "use_fash_attention": false,
  "use_rope": false
}