Update 7-inverse-design.ipynb

docs/source/tutorials/7-inverse-design.ipynb

@@ -1330,7 +1330,7 @@
     "* https://github.com/yoshida-lab/XenonPy/releases/download/v0.4.1/ngram_pubchem_ikebata_reO15_O10.tar.gz\n",
     "* https://github.com/yoshida-lab/XenonPy/releases/download/v0.4.1/ngram_pubchem_ikebata_reO15_O11to20.tar.gz\n",
     "\n",
-    "Please unzip the files yourself.\n",
+    "Please unzip the files yourself. (Currently, you need to manually add the "_sample_order" attribute as well, e.g., setting it to be (0,10).)\n",
     "\n",
     "Using the merge table function in `NGram`, we can train and store the big `NGram` tables separately, and combine them later, as will be demonstrated below. Note that the most efficient way to manually parallelize the training of `NGram` is NOT to split the order, but to split the training set of SMILES strings, which is not what we are doing here."
    ]