Adding filter to work on more

HuggingFace_Mistral_Transformer_Single_Instrument.ipynb

@@ -185,6 +185,38 @@
     "kwargs_dataset = {\"max_seq_len\": sequence_length, \"tokenizer\": tokenizer, \"bos_token_id\": tokenizer[\"BOS_None\"], \"eos_token_id\": tokenizer[\"EOS_None\"]}"
    ]
   },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def remove_files_with_boring_data( file_paths: list[Path], rms_threshold: float = 0.01) -> list[Path]:\n",
+    "    \"\"\"\n",
+    "    Remove files with boring data, i.e. files with low RMS.\n",
+    "    \"\"\"\n",
+    "    from symusic import Score\n",
+    "    from tqdm import tqdm\n",
+    "    import numpy as np\n",
+    "\n",
+    "    rms = lambda data: (sum(x * x for x in data) / len(data)) ** 0.5\n",
+    "\n",
+    "    filtered_files = []\n",
+    "    for file_path in tqdm(file_paths, desc=\"Filtering boring files\"):\n",
+    "        try:\n",
+    "            scores = [Score(file_path)]\n",
+    "        except SCORE_LOADING_EXCEPTION:\n",
+    "            continue\n",
+    "\n",
+    "        for track in scores[0].tracks:\n",
+    "            values = track.notes['pitch']\n",
+    "            result = rms(values)\n",
+    "\n",
+    "\n",
+    "            filtered_files.append(file_path)\n",
+    "    return filtered_files"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,