Обновить библиотеку, читалку и выпуск до 2.31

2026-07-17 14:38:49 +03:00
parent 2f40569031
commit c880395341
188 changed files with 15734 additions and 3702 deletions
@@ -0,0 +1,65 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+if [[ $# -lt 3 ]]; then
+  echo "Usage: $0 DATASET_DIR OUTPUT_DIR PIPER_REPO [EPOCHS]" >&2
+  exit 2
+fi
+
+dataset_dir="$(realpath "$1")"
+output_dir="$(mkdir -p "$2" && realpath "$2")"
+piper_repo="$(realpath "$3")"
+epochs="${4:-2000}"
+
+metadata="$dataset_dir/metadata.csv"
+audio_dir="$dataset_dir/wav"
+if [[ ! -f "$metadata" || ! -d "$audio_dir" ]]; then
+  echo "Dataset must contain metadata.csv and wav/" >&2
+  exit 2
+fi
+if [[ ! -f "$piper_repo/src/piper/train/__main__.py" ]]; then
+  echo "Piper training sources not found at $piper_repo" >&2
+  exit 2
+fi
+
+python3 -m piper.train fit \
+  --data.voice_name aletheia_ru \
+  --data.csv_path "$metadata" \
+  --data.audio_dir "$audio_dir" \
+  --data.espeak_voice ru \
+  --data.cache_dir "$output_dir/cache" \
+  --data.config_path "$output_dir/aletheia_ru.onnx.json" \
+  --data.batch_size 16 \
+  --data.num_workers 4 \
+  --model.sample_rate 22050 \
+  --trainer.accelerator gpu \
+  --trainer.devices 1 \
+  --trainer.precision 16-mixed \
+  --trainer.max_epochs "$epochs" \
+  --trainer.default_root_dir "$output_dir/checkpoints"
+
+checkpoint="$(find "$output_dir/checkpoints" -type f -name '*.ckpt' -printf '%T@ %p\n' | sort -nr | head -n 1 | cut -d' ' -f2-)"
+if [[ -z "$checkpoint" ]]; then
+  echo "Training finished without a checkpoint" >&2
+  exit 1
+fi
+
+python3 -m piper.train.export_onnx \
+  --checkpoint "$checkpoint" \
+  --output-file "$output_dir/aletheia_ru.onnx"
+
+python3 - "$output_dir" <<'PY'
+import hashlib
+import json
+import pathlib
+import sys
+
+root = pathlib.Path(sys.argv[1])
+artifacts = {}
+for name in ("aletheia_ru.onnx", "aletheia_ru.onnx.json"):
+    path = root / name
+    digest = hashlib.sha256(path.read_bytes()).hexdigest()
+    artifacts[name] = {"bytes": path.stat().st_size, "sha256": digest}
+(root / "artifacts.json").write_text(json.dumps(artifacts, indent=2), encoding="utf-8")
+print(json.dumps(artifacts, indent=2))
+PY