| {"MIN_DIFFICULTY": 2, "MAX_DIFFICULTY": 6, "TRAINING_SAMPLES": 3000000, "CONTEXT_LENGTH": 256, "RESIDUAL_EMBEDDING_SIZE": 512, "MLP_EMBEDDING_SIZE": 2048, "NUM_ATTENTION_HEADS": 4, "NUM_LAYERS": 12, "VOCAB_SIZE": 40, "TOTAL_TOKENS": 1214656000, "EPOCHS": 2} |