{"MIN_DIFFICULTY": 2, "MAX_DIFFICULTY": 4, "TRAINING_SAMPLES": 3000000, "CONTEXT_LENGTH": 256, "RESIDUAL_EMBEDDING_SIZE": 256, "MLP_EMBEDDING_SIZE": 1024, "NUM_ATTENTION_HEADS": 4, "NUM_LAYERS": 8, "VOCAB_SIZE": 40, "TOTAL_TOKENS": 419612160}