fix: fix indextts2 model resource checks (#707)

* fork/PR_CLI: 修复CLI误将pinyin.vocab作为模型文件进行检查而报错的问题

* fork/PR_CLI: indextts2 check 指令增加 hf_cache 相关模型的检查

cli_tests/test_cli_v2_batch.py

@@ -15,13 +15,21 @@ REQUIRED_MODEL_FILES = [
     "gpt.pth",
     "s2mel.pth",
     "wav2vec2bert_stats.pt",
-    "pinyin.vocab",
     "feat1.pt",
     "feat2.pt",
 ]
 REQUIRED_MODEL_DIRS = [
     "qwen0.6bemo4-merge",
 ]
+AUX_MODEL_FILES = [
+    "hf_cache/semantic_codec_model.safetensors",
+    "hf_cache/campplus_cn_common.bin",
+    "hf_cache/bigvgan/config.json",
+    "hf_cache/bigvgan/bigvgan_generator.pt",
+]
+AUX_MODEL_DIRS = [
+    "hf_cache/w2v-bert-2.0",
+]
 
 
 def make_model_dir(base_dir):
@@ -31,6 +39,14 @@ def make_model_dir(base_dir):
         (model_dir / filename).write_text("placeholder", encoding="utf-8")
     for dirname in REQUIRED_MODEL_DIRS:
         (model_dir / dirname).mkdir()
+    for filename in AUX_MODEL_FILES:
+        target = model_dir / filename
+        target.parent.mkdir(parents=True, exist_ok=True)
+        target.write_text("placeholder", encoding="utf-8")
+    for dirname in AUX_MODEL_DIRS:
+        target = model_dir / dirname
+        target.mkdir(parents=True, exist_ok=True)
+        (target / "config.json").write_text("placeholder", encoding="utf-8")
     return model_dir
 
 

cli_tests/test_cli_v2_check.py

@@ -16,25 +16,46 @@ REQUIRED_MODEL_FILES = [
     "gpt.pth",
     "s2mel.pth",
     "wav2vec2bert_stats.pt",
-    "pinyin.vocab",
     "feat1.pt",
     "feat2.pt",
 ]
 REQUIRED_MODEL_DIRS = [
     "qwen0.6bemo4-merge",
 ]
+AUX_MODEL_FILES = [
+    "hf_cache/semantic_codec_model.safetensors",
+    "hf_cache/campplus_cn_common.bin",
+    "hf_cache/bigvgan/config.json",
+    "hf_cache/bigvgan/bigvgan_generator.pt",
+]
+AUX_MODEL_DIRS = [
+    "hf_cache/w2v-bert-2.0",
+]
 
 
-def make_model_dir(base_dir):
+def make_model_dir(base_dir, include_aux=True):
     model_dir = base_dir / "checkpoints"
     model_dir.mkdir()
     for filename in REQUIRED_MODEL_FILES:
         (model_dir / filename).write_text("placeholder", encoding="utf-8")
     for dirname in REQUIRED_MODEL_DIRS:
         (model_dir / dirname).mkdir()
+    if include_aux:
+        make_aux_model_cache(model_dir)
     return model_dir
 
 
+def make_aux_model_cache(model_dir):
+    for filename in AUX_MODEL_FILES:
+        target = model_dir / filename
+        target.parent.mkdir(parents=True, exist_ok=True)
+        target.write_text("placeholder", encoding="utf-8")
+    for dirname in AUX_MODEL_DIRS:
+        target = model_dir / dirname
+        target.mkdir(parents=True, exist_ok=True)
+        (target / "config.json").write_text("placeholder", encoding="utf-8")
+
+
 def assert_model_resource_help(test_case, stderr, model_dir):
     test_case.assertIn(f"Model directory: {model_dir}", stderr)
     test_case.assertIn("Missing resources:", stderr)
@@ -189,11 +210,30 @@ class CheckCommandTests(unittest.TestCase):
 
             self.assertEqual(exit_code, 2)
             self.assertEqual(stdout.getvalue(), "")
-            self.assertIn("pinyin.vocab", stderr.getvalue())
             self.assertIn("feat1.pt", stderr.getvalue())
             self.assertIn("feat2.pt", stderr.getvalue())
             self.assertIn("qwen0.6bemo4-merge", stderr.getvalue())
 
+    def test_check_requires_the_auxiliary_model_cache_resources(self):
+        with tempfile.TemporaryDirectory() as temp_dir:
+            model_dir = make_model_dir(Path(temp_dir), include_aux=False)
+
+            from indextts.cli_v2 import main
+
+            stdout = io.StringIO()
+            stderr = io.StringIO()
+            with contextlib.redirect_stdout(stdout), contextlib.redirect_stderr(stderr):
+                exit_code = main(["check", "--model-dir", str(model_dir)])
+
+            self.assertEqual(exit_code, 2)
+            self.assertEqual(stdout.getvalue(), "")
+            self.assertIn("ERROR: missing required model files", stderr.getvalue())
+            self.assertIn("hf_cache/w2v-bert-2.0", stderr.getvalue())
+            self.assertIn("hf_cache/semantic_codec_model.safetensors", stderr.getvalue())
+            self.assertIn("hf_cache/campplus_cn_common.bin", stderr.getvalue())
+            self.assertIn("hf_cache/bigvgan/config.json", stderr.getvalue())
+            self.assertIn("hf_cache/bigvgan/bigvgan_generator.pt", stderr.getvalue())
+
     def test_check_requires_file_resources_and_directory_resources(self):
         with tempfile.TemporaryDirectory() as temp_dir:
             model_dir = Path(temp_dir) / "checkpoints"
@@ -1114,14 +1154,9 @@ class SynthCommandTests(unittest.TestCase):
     def test_synth_maps_runtime_options_to_indextts2(self):
         with tempfile.TemporaryDirectory() as temp_dir:
             temp_path = Path(temp_dir)
-            model_dir = temp_path / "models"
+            model_dir = make_model_dir(temp_path)
             voice_path = temp_path / "voice.wav"
             output_path = temp_path / "out.wav"
-            model_dir.mkdir()
-            for filename in REQUIRED_MODEL_FILES:
-                (model_dir / filename).write_text("placeholder", encoding="utf-8")
-            for dirname in REQUIRED_MODEL_DIRS:
-                (model_dir / dirname).mkdir()
             voice_path.write_bytes(b"voice")
 
             exit_code, stdout, stderr, calls = self.run_synth(

cli_tests/test_cli_v2_config.py

@@ -16,13 +16,21 @@ REQUIRED_MODEL_FILES = [
     "gpt.pth",
     "s2mel.pth",
     "wav2vec2bert_stats.pt",
-    "pinyin.vocab",
     "feat1.pt",
     "feat2.pt",
 ]
 REQUIRED_MODEL_DIRS = [
     "qwen0.6bemo4-merge",
 ]
+AUX_MODEL_FILES = [
+    "hf_cache/semantic_codec_model.safetensors",
+    "hf_cache/campplus_cn_common.bin",
+    "hf_cache/bigvgan/config.json",
+    "hf_cache/bigvgan/bigvgan_generator.pt",
+]
+AUX_MODEL_DIRS = [
+    "hf_cache/w2v-bert-2.0",
+]
 
 
 def make_model_dir(path):
@@ -31,6 +39,14 @@ def make_model_dir(path):
         (path / filename).write_text("placeholder", encoding="utf-8")
     for dirname in REQUIRED_MODEL_DIRS:
         (path / dirname).mkdir()
+    for filename in AUX_MODEL_FILES:
+        target = path / filename
+        target.parent.mkdir(parents=True, exist_ok=True)
+        target.write_text("placeholder", encoding="utf-8")
+    for dirname in AUX_MODEL_DIRS:
+        target = path / dirname
+        target.mkdir(parents=True, exist_ok=True)
+        (target / "config.json").write_text("placeholder", encoding="utf-8")
 
 
 def fake_torch():

cli_tests/test_cli_v2_download.py

@@ -14,21 +14,42 @@ REQUIRED_MODEL_FILES = [
     "gpt.pth",
     "s2mel.pth",
     "wav2vec2bert_stats.pt",
-    "pinyin.vocab",
     "feat1.pt",
     "feat2.pt",
 ]
 REQUIRED_MODEL_DIRS = [
     "qwen0.6bemo4-merge",
 ]
+AUX_MODEL_FILES = [
+    "hf_cache/semantic_codec_model.safetensors",
+    "hf_cache/campplus_cn_common.bin",
+    "hf_cache/bigvgan/config.json",
+    "hf_cache/bigvgan/bigvgan_generator.pt",
+]
+AUX_MODEL_DIRS = [
+    "hf_cache/w2v-bert-2.0",
+]
 
 
-def make_model_dir(path):
+def make_model_dir(path, include_aux=True):
     path.mkdir(parents=True, exist_ok=True)
     for filename in REQUIRED_MODEL_FILES:
         (path / filename).write_text("placeholder", encoding="utf-8")
     for dirname in REQUIRED_MODEL_DIRS:
         (path / dirname).mkdir(exist_ok=True)
+    if include_aux:
+        make_aux_model_cache(path)
+
+
+def make_aux_model_cache(path):
+    for filename in AUX_MODEL_FILES:
+        target = path / filename
+        target.parent.mkdir(parents=True, exist_ok=True)
+        target.write_text("placeholder", encoding="utf-8")
+    for dirname in AUX_MODEL_DIRS:
+        target = path / dirname
+        target.mkdir(parents=True, exist_ok=True)
+        (target / "config.json").write_text("placeholder", encoding="utf-8")
 
 
 def user_state_paths(temp_path):
@@ -72,19 +93,29 @@ class DownloadCommandTests(unittest.TestCase):
         with tempfile.TemporaryDirectory() as temp_dir:
             state = user_state_paths(Path(temp_dir).resolve())
             calls = []
+            aux_calls = []
 
             def fake_snapshot_download(repo_id, local_dir, **kwargs):
                 calls.append((repo_id, Path(local_dir)))
-                make_model_dir(Path(local_dir))
+                make_model_dir(Path(local_dir), include_aux=False)
                 return str(local_dir)
 
+            def fake_ensure_models_available(model_dir):
+                aux_calls.append(Path(model_dir))
+                make_aux_model_cache(Path(model_dir))
+
             with mock.patch.dict(os.environ, state["env"], clear=False):
                 with mock.patch("indextts.utils.model_download.snapshot_download", side_effect=fake_snapshot_download):
-                    exit_code, stdout, stderr = self.run_cli(["download"])
+                    with mock.patch(
+                        "indextts.utils.model_download.ensure_models_available",
+                        side_effect=fake_ensure_models_available,
+                    ):
+                        exit_code, stdout, stderr = self.run_cli(["download"])
                 config_exists = state["config_path"].exists()
 
         self.assertEqual(exit_code, 0)
         self.assertEqual(calls, [("IndexTeam/IndexTTS-2", state["model_dir"])])
+        self.assertEqual(aux_calls, [state["model_dir"]])
         self.assertIn(f"Downloaded model resources to: {state['model_dir']}", stdout)
         self.assertEqual(stderr, "")
         self.assertFalse(config_exists)

indextts/cli_v2.py

@@ -25,13 +25,21 @@ REQUIRED_MODEL_FILES = (
     "gpt.pth",
     "s2mel.pth",
     "wav2vec2bert_stats.pt",
-    "pinyin.vocab",
     "feat1.pt",
     "feat2.pt",
 )
 REQUIRED_MODEL_DIRS = (
     "qwen0.6bemo4-merge",
 )
+REQUIRED_AUX_MODEL_FILES = (
+    "hf_cache/semantic_codec_model.safetensors",
+    "hf_cache/campplus_cn_common.bin",
+    "hf_cache/bigvgan/config.json",
+    "hf_cache/bigvgan/bigvgan_generator.pt",
+)
+REQUIRED_AUX_MODEL_DIRS = (
+    "hf_cache/w2v-bert-2.0",
+)
 MODEL_REPO_ID = "IndexTeam/IndexTTS-2"
 REQUIRED_PACKAGES = ("torch", "torchaudio", "indextts")
 PERSISTED_CONFIG_KEYS = (
@@ -322,6 +330,13 @@ def _download_model_resources(source, model_dir):
         from huggingface_hub import snapshot_download
         snapshot_download(repo_id=MODEL_REPO_ID, local_dir=str(model_dir))
 
+    if _missing_primary_model_resources(model_dir):
+        return
+
+    from indextts.utils.model_download import ensure_models_available
+
+    ensure_models_available(str(model_dir))
+
 
 def _download_support_package(source):
     if source == "auto":
@@ -1517,11 +1532,26 @@ def _print_model_resource_help(model_dir, missing_summary):
 def _missing_model_files(model_dir):
     if not model_dir.is_dir():
         return None
+    missing_files = _missing_primary_model_resources(model_dir)
+    missing_aux_files = [
+        filename for filename in REQUIRED_AUX_MODEL_FILES if not _model_resource_path(model_dir, filename).is_file()
+    ]
+    missing_aux_dirs = [
+        dirname for dirname in REQUIRED_AUX_MODEL_DIRS if not _model_resource_path(model_dir, dirname).is_dir()
+    ]
+    return missing_files + missing_aux_files + missing_aux_dirs
+
+
+def _missing_primary_model_resources(model_dir):
     missing_files = [filename for filename in REQUIRED_MODEL_FILES if not (model_dir / filename).is_file()]
     missing_dirs = [dirname for dirname in REQUIRED_MODEL_DIRS if not (model_dir / dirname).is_dir()]
     return missing_files + missing_dirs
 
 
+def _model_resource_path(model_dir, relative_path):
+    return model_dir.joinpath(*relative_path.split("/"))
+
+
 def _import_required_packages():
     missing = []
     imported = {}