feat(scripts): TCP server setup add PyTorch nightly install

scripts/setup/setup-python-dev-env.js

@@ -130,6 +130,30 @@ SPACY_MODELS.set('fr', {
   const pipfileMtime = fs.statSync(pipfilePath).mtime
   const hasDotVenv = fs.existsSync(dotVenvPath)
   const { type: osType, cpuArchitecture } = SystemHelper.getInformation()
+  /**
+   * Install PyTorch nightly to support CUDA 12.4
+   * as it is required by the latest NVIDIA drivers for CUDA runtime APIs
+   *
+   * @see https://stackoverflow.com/a/76972265/1768162
+   */
+  const installPytorch = async () => {
+    LogHelper.info('Installing PyTorch nightly with CUDA support...')
+    try {
+      await command(
+        'pipenv run pip install --pre torch --index-url https://download.pytorch.org/whl/nightly/cu124',
+        {
+          shell: true,
+          stdio: 'inherit'
+        }
+      )
+      LogHelper.success('PyTorch nightly with CUDA support installed')
+    } catch (e) {
+      LogHelper.error(
+        `Failed to install PyTorch nightly with CUDA support: ${e}`
+      )
+      process.exit(1)
+    }
+  }
   const installPythonPackages = async () => {
     LogHelper.info(`Installing Python packages from ${pipfilePath}.lock...`)
 
@@ -177,6 +201,8 @@ SPACY_MODELS.set('fr', {
       }
 
       LogHelper.success('Python packages installed')
+
+      await installPytorch()
     } catch (e) {
       LogHelper.error(`Failed to install Python packages: ${e}`)
 

tcp_server/src/Pipfile

@@ -22,7 +22,8 @@ spacy = "==3.5.4"
 geonamescache = "==1.6.0"
 
 # TCP server; TTS
-torch = "==1.12.1"
+# PyTorch is installed via the setup script
+# torch = "*"
 
 # TTS
 transformers = "==4.27.4"

tcp_server/src/lib/tts/utils.py

@@ -32,7 +32,6 @@ def get_text_for_tts_infer(text, language_str, hps, device, symbol_to_id=None):
         ja_bert = torch.zeros(768, len(phone))
     else:
         bert = get_bert(norm_text, word2ph, language_str, device)
-        print('bert', bert)
         del word2ph
         assert bert.shape[-1] == len(phone), phone
 

tcp_server/src/lib/tts_to_delete/tts.py

@@ -1,75 +0,0 @@
-import os
-import torch
-
-from . import utils
-from ..constants import TTS_MODEL_CONFIG_PATH, TTS_MODEL_PATH, IS_TTS_ENABLED
-
-class TTS:
-    def __init__(self):
-        self.hyper_params = None
-        self.device = 'auto'
-        self.num_languages = None
-        self.num_tones = None
-        self.symbols = None
-
-        if not IS_TTS_ENABLED:
-            self.log('TTS is disabled')
-            return
-
-        self.log('Loading model...')
-
-        if not self.has_model_config():
-            self.log(f'Model config not found at {TTS_MODEL_CONFIG_PATH}')
-            return
-
-        if not self.is_model_downloaded():
-            self.log(f'Model not found at {TTS_MODEL_PATH}')
-            return
-
-        self.set_device()
-        self.hyper_params = utils.get_hparams_from_file(TTS_MODEL_CONFIG_PATH)
-        self.num_languages = self.hyper_params.num_languages
-        self.num_tones = self.hyper_params.num_tones
-        self.symbols = self.hyper_params.symbols
-
-        model = SynthesizerTrn(
-            len(self.symbols),
-            self.hyper_params.data.filter_length // 2 + 1,
-            self.hyper_params.train.segment_size // self.hyper_params.data.hop_length,
-            n_speakers=self.hyper_params.data.n_speakers,
-            num_tones=self.num_tones,
-            num_languages=self.num_languages,
-            **self.hyper_params.model,
-        ).to(self.device)
-        model.eval()
-        self.model = model
-
-        self.log('Model loaded')
-
-    def set_device(self):
-        if self.device == 'auto':
-            self.device = 'cpu'
-
-            if torch.cuda.is_available():
-                self.device = 'cuda'
-            else:
-                self.log('GPU not available. CUDA is not installed?')
-
-            if torch.backends.mps.is_available():
-                self.device = 'mps'
-        if 'cuda' in self.device:
-            assert torch.cuda.is_available()
-
-        self.log(f'Device: {self.device}')
-
-    @staticmethod
-    def is_model_downloaded():
-        return os.path.exists(TTS_MODEL_PATH)
-
-    @staticmethod
-    def has_model_config():
-        return os.path.exists(TTS_MODEL_CONFIG_PATH)
-
-    @staticmethod
-    def log(*args, **kwargs):
-        print('[TTS]', *args, **kwargs)

tcp_server/src/lib/tts_to_delete/utils.py

@@ -1,40 +0,0 @@
-import json
-
-def get_hparams_from_file(config_path):
-    with open(config_path, "r", encoding="utf-8") as f:
-        data = f.read()
-    config = json.loads(data)
-
-    hparams = HParams(**config)
-    return hparams
-
-class HParams:
-    def __init__(self, **kwargs):
-        for k, v in kwargs.items():
-            if type(v) == dict:
-                v = HParams(**v)
-            self[k] = v
-
-    def keys(self):
-        return self.__dict__.keys()
-
-    def items(self):
-        return self.__dict__.items()
-
-    def values(self):
-        return self.__dict__.values()
-
-    def __len__(self):
-        return len(self.__dict__)
-
-    def __getitem__(self, key):
-        return getattr(self, key)
-
-    def __setitem__(self, key, value):
-        return setattr(self, key, value)
-
-    def __contains__(self, key):
-        return key in self.__dict__
-
-    def __repr__(self):
-        return self.__dict__.__repr__()