diff --git a/.DS_Store b/.DS_Store new file mode 100644 index 0000000..5008ddf Binary files /dev/null and b/.DS_Store differ diff --git a/microWakeWord_training_notebook.ipynb b/microWakeWord_training_notebook.ipynb index 5fe2110..b848e8b 100644 --- a/microWakeWord_training_notebook.ipynb +++ b/microWakeWord_training_notebook.ipynb @@ -232,7 +232,7 @@ }, "outputs": [], "source": [ - "# Generate a large number of wake word samples for training\n", + "# Generate a large number of wake word samples for training (with length-scale sweep)\n", "import sys, subprocess\n", "from pathlib import Path\n", "\n", @@ -240,14 +240,21 @@ "MODELS_DIR = REPO_DIR / \"models\"\n", "MODEL_NAME = \"en_US-libritts_r-medium.pt\"\n", "\n", + "MAX_SAMPLES = 50000\n", + "BATCH_SIZE = 100\n", + "\n", + "# Piper \"speed\" control via piper-sample-generator is length_scale(s)\n", + "LENGTH_SCALES = [\"0.85\", \"0.95\", \"1.00\", \"1.05\", \"1.15\"]\n", + "\n", "cmd = [\n", " sys.executable,\n", " str(REPO_DIR / \"generate_samples.py\"),\n", " TARGET_WORD,\n", " \"--model\", str(MODELS_DIR / MODEL_NAME),\n", - " \"--max-samples\", \"50000\",\n", - " \"--batch-size\", \"100\",\n", + " \"--max-samples\", str(MAX_SAMPLES),\n", + " \"--batch-size\", str(BATCH_SIZE),\n", " \"--output-dir\", \"generated_samples\",\n", + " \"--length-scales\", *LENGTH_SCALES,\n", "]\n", "\n", "print(\"→\", \" \".join(cmd))\n",