{
  "copied_tensor_count": 777,
  "converted_linear_count": 410,
  "embedding_factorized": true,
  "embedding_rank": 16,
  "out_dtype": "torch.bfloat16"
}