Update advanced_training_notebook.ipynb

advanced_training_notebook.ipynb

@@ -174,8 +174,7 @@
     "# Download and Process Audioset\n",
     "# -----------------------------\n",
     "audioset_dir = \"./audioset\"\n",
-    "if not os.path.exists(audioset_dir):\n",
+    "os.makedirs(audioset_dir, exist_ok=True)\n",
-    "    os.mkdir(audioset_dir)\n",
     "\n",
     "# Full-scale dataset download links\n",
     "dataset_links = [\n",
@@ -183,46 +182,56 @@
     "    for i in range(10)\n",
     "]\n",
     "\n",
+    "# Download and extract each dataset part\n",
     "for link in dataset_links:\n",
     "    file_name = link.split(\"/\")[-1]\n",
-    "    out_dir = os.path.join(audioset_dir, file_name)\n",
+    "    out_path = os.path.join(audioset_dir, file_name)\n",
-    "    if not os.path.exists(out_dir):\n",
+    "    if not os.path.exists(out_path):\n",
     "        print(f\"Downloading {file_name}...\")\n",
-    "        os.system(f\"wget --quiet -O {out_dir} {link}\")\n",
+    "        os.system(f\"wget --quiet -O {out_path} {link}\")\n",
     "        print(f\"Extracting {file_name}...\")\n",
-    "        os.system(f\"cd {audioset_dir} && tar -xf {file_name}\")\n",
+    "        os.system(f\"tar -xf {out_path} -C {audioset_dir}\")\n",
     "\n",
+    "# Directory for 16kHz WAV files\n",
     "output_dir = \"./audioset_16k\"\n",
-    "if not os.path.exists(output_dir):\n",
+    "os.makedirs(output_dir, exist_ok=True)\n",
-    "    os.mkdir(output_dir)\n",
     "\n",
-    "# Save clips to 16-bit PCM wav files\n",
+    "# Collect all FLAC files for processing\n",
-    "audioset_files = list(Path(\"audioset/audio\").glob(\"**/*.flac\"))\n",
+    "audioset_files = list(Path(audioset_dir).glob(\"**/*.flac\"))\n",
     "print(f\"Number of FLAC files found: {len(audioset_files)}\")\n",
+    "\n",
     "if audioset_files:\n",
     "    audioset_dataset = Dataset.from_dict({\"audio\": [str(file) for file in audioset_files]})\n",
     "    audioset_dataset = audioset_dataset.cast_column(\"audio\", Audio(sampling_rate=16000))\n",
     "\n",
     "    corrupted_files = []\n",
     "    print(\"Converting Audioset files to 16kHz WAV...\")\n",
-    "    for row in tqdm(audioset_dataset):\n",
+    "\n",
+    "    for row in tqdm(audioset_dataset, desc=\"Processing Audioset files\"):\n",
     "        try:\n",
-    "            name = row[\"audio\"][\"path\"].split(\"/\")[-1].replace(\".flac\", \".wav\")\n",
+    "            # Define output file name and path\n",
-    "            scipy.io.wavfile.write(\n",
+    "            name = Path(row[\"audio\"][\"path\"]).stem + \".wav\"\n",
-    "                os.path.join(output_dir, name), \n",
+    "            output_path = Path(output_dir) / name\n",
-    "                16000, \n",
+    "            \n",
-    "                (row[\"audio\"][\"array\"] * 32767).astype(np.int16)\n",
+    "            # Read and convert audio\n",
-    "            )\n",
+    "            data = row[\"audio\"][\"array\"]\n",
+    "            scipy.io.wavfile.write(output_path, 16000, (data * 32767).astype(np.int16))\n",
+    "        \n",
     "        except Exception as e:\n",
     "            print(f\"Error converting {row['audio']['path']}: {e}\")\n",
     "            corrupted_files.append(row[\"audio\"][\"path\"])\n",
     "\n",
+    "    # Log corrupted files\n",
     "    if corrupted_files:\n",
-    "        with open(\"audioset_corrupted_files.log\", \"w\") as log_file:\n",
+    "        log_path = Path(output_dir) / \"audioset_corrupted_files.log\"\n",
+    "        with open(log_path, \"w\") as log_file:\n",
     "            log_file.writelines(f\"{file}\\n\" for file in corrupted_files)\n",
+    "        print(f\"Logged corrupted files to {log_path}\")\n",
     "else:\n",
     "    print(\"No FLAC files found in Audioset.\")\n",
     "\n",
+    "print(\"Audioset processing complete!\")\n",
+    "\n",
     "# -----------------------------\n",
     "# Download and Process FMA\n",
     "# -----------------------------\n",