weights/model_config.json

{
  "vocab_size": 8000,
  "d_model": 256,
  "num_heads": 8,
  "num_layers": 6,
  "d_ff": 1024,
  "max_seq_length": 512,
  "dropout": 0.1,
  "batch_size": 8,
  "learning_rate": 0.0001,
  "num_epochs": 5,
  "gradient_clip": 1.0,
  "warmup_steps": 500,
  "dataset_name": "institutionai/institutional-books-1.0",
  "filter_years": [
    1800,
    1950
  ],
  "target_subjects": [
    "physics",
    "quantum",
    "medicine",
    "medical",
    "biology",
    "disease",
    "cure",
    "therapy",
    "anatomy",
    "chemistry",
    "cancer",
    "cell",
    "darwin",
    "evolution",
    "relativity",
    "electromagnetic",
    "radiation",
    "molecular"
  ],
  "max_books": 200,
  "max_text_length": 100000,
  "num_adapters_target": 100,
  "tcl_compression_ratio": 0.1,
  "output_dir": "./jarvis_v1_oracle",
  "checkpoint_every": 500
}