Merge pull request #1265 from coqui-ai/dev-fix-glowtts-infer

Fix multi-speaker GlowTTS inference

.github/workflows/inference_tests.yml

@@ -0,0 +1,46 @@
+name: inference_tests
+
+on:
+  push:
+    branches:
+      - main
+  pull_request:
+    types: [opened, synchronize, reopened]
+jobs:
+  check_skip:
+    runs-on: ubuntu-latest
+    if: "! contains(github.event.head_commit.message, '[ci skip]')"
+    steps:
+      - run: echo "${{ github.event.head_commit.message }}"
+
+  test:
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        python-version: [3.6, 3.7, 3.8, 3.9]
+        experimental: [false]
+    steps:
+      - uses: actions/checkout@v2
+      - name: Set up Python ${{ matrix.python-version }}
+        uses: coqui-ai/setup-python@pip-cache-key-py-ver
+        with:
+          python-version: ${{ matrix.python-version }}
+          architecture: x64
+          cache: 'pip'
+          cache-dependency-path: 'requirements*'
+      - name: check OS
+        run: cat /etc/os-release
+      - name: Install dependencies
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y --no-install-recommends git make gcc
+          make system-deps
+      - name: Install/upgrade Python setup deps
+        run: python3 -m pip install --upgrade pip setuptools wheel
+      - name: Install TTS
+        run: |
+          python3 -m pip install .[all]
+          python3 setup.py egg_info
+      - name: Unit tests
+        run: make inference_tests

Makefile

@@ -26,6 +26,9 @@ test_aux:	## run aux tests.
 test_zoo:	## run zoo tests.
 	nosetests tests.zoo_tests -x --with-cov -cov  --cover-erase --cover-package TTS tests.zoo_tests --nologcapture --with-id
 
+inference_tests: ## run inference tests.
+	nosetests tests.inference_tests -x --with-cov -cov  --cover-erase --cover-package TTS tests.inference_tests --nologcapture --with-id	
+
 data_tests:	## run data tests.
 	nosetests tests.data_tests -x --with-cov -cov  --cover-erase --cover-package TTS tests.data_tests --nologcapture --with-id
 

TTS/bin/extract_tts_spectrograms.py

@@ -138,7 +138,7 @@ def inference(
             aux_input={"d_vectors": speaker_c, "speaker_ids": speaker_ids},
         )
         model_output = outputs["model_outputs"]
-        model_output = model_output.transpose(1, 2).detach().cpu().numpy()
+        model_output = model_output.detach().cpu().numpy()
 
     elif "tacotron" in model_name:
         aux_input = {"speaker_ids": speaker_ids, "d_vectors": d_vectors}

TTS/tts/models/glow_tts.py

@@ -170,6 +170,8 @@ class GlowTTS(BaseTTS):
         if g is not None:
             if hasattr(self, "emb_g"):
                 # use speaker embedding layer
+                if not g.size(): # if is a scalar
+                    g = g.unsqueeze(0) # unsqueeze
                 g = F.normalize(self.emb_g(g)).unsqueeze(-1)  # [b, h, 1]
             else:
                 # use d-vector

tests/inference_tests/test_synthesizer.py

@@ -6,7 +6,7 @@ from TTS.tts.models import setup_model
 from TTS.utils.io import save_checkpoint
 from TTS.utils.synthesizer import Synthesizer
 
-from .. import get_tests_output_path
+from tests import get_tests_output_path
 
 
 class SynthesizerTest(unittest.TestCase):

tests/tts_tests/test_align_tts_train.py

@@ -4,6 +4,7 @@ import shutil
 
 from tests import get_device_id, get_tests_output_path, run_cli
 from TTS.tts.configs.align_tts_config import AlignTTSConfig
+from TTS.utils.trainer_utils import get_last_checkpoint
 
 config_path = os.path.join(get_tests_output_path(), "test_model_config.json")
 output_path = os.path.join(get_tests_output_path(), "train_outputs")
@@ -47,6 +48,14 @@ run_cli(command_train)
 # Find latest folder
 continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
 
+# Inference using TTS API
+continue_config_path = os.path.join(continue_path, "config.json")
+continue_restore_path, _ = get_last_checkpoint(continue_path)
+out_wav_path = os.path.join(get_tests_output_path(), 'output.wav')
+
+inference_command = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' tts --text 'This is an example.' --config_path {continue_config_path} --model_path {continue_restore_path} --out_path {out_wav_path}"
+run_cli(inference_command)
+
 # restore the model and continue training for one more epoch
 command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_tts.py --continue_path {continue_path} "
 run_cli(command_train)

tests/tts_tests/test_fast_pitch_speaker_emb_train.py

@@ -0,0 +1,82 @@
+import glob
+import os
+import shutil
+
+from tests import get_device_id, get_tests_output_path, run_cli
+from TTS.config.shared_configs import BaseAudioConfig
+from TTS.tts.configs.fast_pitch_config import FastPitchConfig
+from TTS.utils.trainer_utils import get_last_checkpoint
+
+config_path = os.path.join(get_tests_output_path(), "test_model_config.json")
+output_path = os.path.join(get_tests_output_path(), "train_outputs")
+
+audio_config = BaseAudioConfig(
+    sample_rate=22050,
+    do_trim_silence=True,
+    trim_db=60.0,
+    signal_norm=False,
+    mel_fmin=0.0,
+    mel_fmax=8000,
+    spec_gain=1.0,
+    log_func="np.log",
+    ref_level_db=20,
+    preemphasis=0.0,
+)
+
+config = FastPitchConfig(
+    audio=audio_config,
+    batch_size=8,
+    eval_batch_size=8,
+    num_loader_workers=0,
+    num_eval_loader_workers=0,
+    text_cleaner="english_cleaners",
+    use_phonemes=True,
+    phoneme_language="en-us",
+    phoneme_cache_path="tests/data/ljspeech/phoneme_cache/",
+    f0_cache_path="tests/data/ljspeech/f0_cache/",
+    run_eval=True,
+    test_delay_epochs=-1,
+    epochs=1,
+    print_step=1,
+    print_eval=True,
+    use_speaker_embedding=True,
+    test_sentences=[
+        "Be a voice, not an echo.",
+    ],
+)
+config.audio.do_trim_silence = True
+config.use_speaker_embedding = True
+config.model_args.use_speaker_embedding = True
+config.audio.trim_db = 60
+config.save_json(config_path)
+
+# train the model for one epoch
+command_train = (
+    f"CUDA_VISIBLE_DEVICES='{get_device_id()}'  python TTS/bin/train_tts.py --config_path {config_path}  "
+    f"--coqpit.output_path {output_path} "
+    "--coqpit.datasets.0.name ljspeech_test "
+    "--coqpit.datasets.0.meta_file_train metadata.csv "
+    "--coqpit.datasets.0.meta_file_val metadata.csv "
+    "--coqpit.datasets.0.path tests/data/ljspeech "
+    "--coqpit.datasets.0.meta_file_attn_mask tests/data/ljspeech/metadata_attn_mask.txt "
+    "--coqpit.test_delay_epochs 0"
+)
+run_cli(command_train)
+
+# Find latest folder
+continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
+
+# Inference using TTS API
+continue_config_path = os.path.join(continue_path, "config.json")
+continue_restore_path, _ = get_last_checkpoint(continue_path)
+out_wav_path = os.path.join(get_tests_output_path(), 'output.wav')
+speaker_id = "ljspeech-1"
+continue_speakers_path = os.path.join(continue_path, "speakers.json")
+
+inference_command = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' tts --text 'This is an example.' --speaker_idx {speaker_id} --speakers_file_path {continue_speakers_path} --config_path {continue_config_path} --model_path {continue_restore_path} --out_path {out_wav_path}"
+run_cli(inference_command)
+
+# restore the model and continue training for one more epoch
+command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_tts.py --continue_path {continue_path} "
+run_cli(command_train)
+shutil.rmtree(continue_path)

tests/tts_tests/test_fast_pitch_train.py

@@ -5,8 +5,9 @@ import shutil
 from tests import get_device_id, get_tests_output_path, run_cli
 from TTS.config.shared_configs import BaseAudioConfig
 from TTS.tts.configs.fast_pitch_config import FastPitchConfig
+from TTS.utils.trainer_utils import get_last_checkpoint
 
-config_path = os.path.join(get_tests_output_path(), "test_fast_pitch_config.json")
+config_path = os.path.join(get_tests_output_path(), "test_model_config.json")
 output_path = os.path.join(get_tests_output_path(), "train_outputs")
 
 audio_config = BaseAudioConfig(
@@ -41,8 +42,11 @@ config = FastPitchConfig(
     test_sentences=[
         "Be a voice, not an echo.",
     ],
+    use_speaker_embedding=False,
 )
 config.audio.do_trim_silence = True
+config.use_speaker_embedding = False
+config.model_args.use_speaker_embedding = False
 config.audio.trim_db = 60
 config.save_json(config_path)
 
@@ -57,11 +61,20 @@ command_train = (
     "--coqpit.datasets.0.meta_file_attn_mask tests/data/ljspeech/metadata_attn_mask.txt "
     "--coqpit.test_delay_epochs 0"
 )
+
 run_cli(command_train)
 
 # Find latest folder
 continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
 
+# Inference using TTS API
+continue_config_path = os.path.join(continue_path, "config.json")
+continue_restore_path, _ = get_last_checkpoint(continue_path)
+out_wav_path = os.path.join(get_tests_output_path(), 'output.wav')
+
+inference_command = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' tts --text 'This is an example.' --config_path {continue_config_path} --model_path {continue_restore_path} --out_path {out_wav_path}"
+run_cli(inference_command)
+
 # restore the model and continue training for one more epoch
 command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_tts.py --continue_path {continue_path} "
 run_cli(command_train)

tests/tts_tests/test_glow_tts_d-vectors_train.py

@@ -0,0 +1,70 @@
+import glob
+import os
+import shutil
+
+from tests import get_device_id, get_tests_output_path, run_cli
+from TTS.tts.configs.glow_tts_config import GlowTTSConfig
+from TTS.utils.trainer_utils import get_last_checkpoint
+
+config_path = os.path.join(get_tests_output_path(), "test_model_config.json")
+output_path = os.path.join(get_tests_output_path(), "train_outputs")
+
+
+config = GlowTTSConfig(
+    batch_size=2,
+    eval_batch_size=8,
+    num_loader_workers=0,
+    num_eval_loader_workers=0,
+    text_cleaner="english_cleaners",
+    use_phonemes=True,
+    use_espeak_phonemes=True,
+    phoneme_language="en-us",
+    phoneme_cache_path="tests/data/ljspeech/phoneme_cache/",
+    run_eval=True,
+    test_delay_epochs=-1,
+    epochs=1,
+    print_step=1,
+    print_eval=True,
+    test_sentences=[
+        "Be a voice, not an echo.",
+    ],
+    data_dep_init_steps=1.0,
+    use_speaker_embedding=False,
+    use_d_vector_file=True,
+    d_vector_file="tests/data/ljspeech/speakers.json",
+    d_vector_dim=256,
+)
+config.audio.do_trim_silence = True
+config.audio.trim_db = 60
+config.save_json(config_path)
+
+# train the model for one epoch
+command_train = (
+    f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_tts.py --config_path {config_path} "
+    f"--coqpit.output_path {output_path} "
+    "--coqpit.datasets.0.name ljspeech_test "
+    "--coqpit.datasets.0.meta_file_train metadata.csv "
+    "--coqpit.datasets.0.meta_file_val metadata.csv "
+    "--coqpit.datasets.0.path tests/data/ljspeech "
+    "--coqpit.datasets.0.meta_file_attn_mask tests/data/ljspeech/metadata_attn_mask.txt "
+    "--coqpit.test_delay_epochs 0"
+)
+run_cli(command_train)
+
+# Find latest folder
+continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
+
+# Inference using TTS API
+continue_config_path = os.path.join(continue_path, "config.json")
+continue_restore_path, _ = get_last_checkpoint(continue_path)
+out_wav_path = os.path.join(get_tests_output_path(), 'output.wav')
+speaker_id = "ljspeech-1"
+continue_speakers_path = config.d_vector_file
+
+inference_command = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' tts --text 'This is an example.' --speaker_idx {speaker_id} --speakers_file_path {continue_speakers_path} --config_path {continue_config_path} --model_path {continue_restore_path} --out_path {out_wav_path}"
+run_cli(inference_command)
+
+# restore the model and continue training for one more epoch
+command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_tts.py --continue_path {continue_path} "
+run_cli(command_train)
+shutil.rmtree(continue_path)

tests/tts_tests/test_glow_tts_speaker_emb_train.py

@@ -0,0 +1,67 @@
+import glob
+import os
+import shutil
+
+from tests import get_device_id, get_tests_output_path, run_cli
+from TTS.tts.configs.glow_tts_config import GlowTTSConfig
+from TTS.utils.trainer_utils import get_last_checkpoint
+
+config_path = os.path.join(get_tests_output_path(), "test_model_config.json")
+output_path = os.path.join(get_tests_output_path(), "train_outputs")
+
+
+config = GlowTTSConfig(
+    batch_size=2,
+    eval_batch_size=8,
+    num_loader_workers=0,
+    num_eval_loader_workers=0,
+    text_cleaner="english_cleaners",
+    use_phonemes=True,
+    use_espeak_phonemes=True,
+    phoneme_language="en-us",
+    phoneme_cache_path="tests/data/ljspeech/phoneme_cache/",
+    run_eval=True,
+    test_delay_epochs=-1,
+    epochs=1,
+    print_step=1,
+    print_eval=True,
+    test_sentences=[
+        "Be a voice, not an echo.",
+    ],
+    data_dep_init_steps=1.0,
+    use_speaker_embedding=True,
+)
+config.audio.do_trim_silence = True
+config.audio.trim_db = 60
+config.save_json(config_path)
+
+# train the model for one epoch
+command_train = (
+    f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_tts.py --config_path {config_path} "
+    f"--coqpit.output_path {output_path} "
+    "--coqpit.datasets.0.name ljspeech_test "
+    "--coqpit.datasets.0.meta_file_train metadata.csv "
+    "--coqpit.datasets.0.meta_file_val metadata.csv "
+    "--coqpit.datasets.0.path tests/data/ljspeech "
+    "--coqpit.datasets.0.meta_file_attn_mask tests/data/ljspeech/metadata_attn_mask.txt "
+    "--coqpit.test_delay_epochs 0"
+)
+run_cli(command_train)
+
+# Find latest folder
+continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
+
+# Inference using TTS API
+continue_config_path = os.path.join(continue_path, "config.json")
+continue_restore_path, _ = get_last_checkpoint(continue_path)
+out_wav_path = os.path.join(get_tests_output_path(), 'output.wav')
+speaker_id = "ljspeech-1"
+continue_speakers_path = os.path.join(continue_path, "speakers.json")
+
+inference_command = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' tts --text 'This is an example.' --speaker_idx {speaker_id} --speakers_file_path {continue_speakers_path} --config_path {continue_config_path} --model_path {continue_restore_path} --out_path {out_wav_path}"
+run_cli(inference_command)
+
+# restore the model and continue training for one more epoch
+command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_tts.py --continue_path {continue_path} "
+run_cli(command_train)
+shutil.rmtree(continue_path)

tests/tts_tests/test_glow_tts_train.py

@@ -4,6 +4,7 @@ import shutil
 
 from tests import get_device_id, get_tests_output_path, run_cli
 from TTS.tts.configs.glow_tts_config import GlowTTSConfig
+from TTS.utils.trainer_utils import get_last_checkpoint
 
 config_path = os.path.join(get_tests_output_path(), "test_model_config.json")
 output_path = os.path.join(get_tests_output_path(), "train_outputs")
@@ -49,6 +50,14 @@ run_cli(command_train)
 # Find latest folder
 continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
 
+# Inference using TTS API
+continue_config_path = os.path.join(continue_path, "config.json")
+continue_restore_path, _ = get_last_checkpoint(continue_path)
+out_wav_path = os.path.join(get_tests_output_path(), 'output.wav')
+
+inference_command = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' tts --text 'This is an example.' --config_path {continue_config_path} --model_path {continue_restore_path} --out_path {out_wav_path}"
+run_cli(inference_command)
+
 # restore the model and continue training for one more epoch
 command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_tts.py --continue_path {continue_path} "
 run_cli(command_train)

tests/tts_tests/test_speedy_speech_train.py

@@ -4,6 +4,7 @@ import shutil
 
 from tests import get_device_id, get_tests_output_path, run_cli
 from TTS.tts.configs.speedy_speech_config import SpeedySpeechConfig
+from TTS.utils.trainer_utils import get_last_checkpoint
 
 config_path = os.path.join(get_tests_output_path(), "test_speedy_speech_config.json")
 output_path = os.path.join(get_tests_output_path(), "train_outputs")
@@ -47,6 +48,14 @@ run_cli(command_train)
 # Find latest folder
 continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
 
+# Inference using TTS API
+continue_config_path = os.path.join(continue_path, "config.json")
+continue_restore_path, _ = get_last_checkpoint(continue_path)
+out_wav_path = os.path.join(get_tests_output_path(), 'output.wav')
+
+inference_command = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' tts --text 'This is an example for it.' --config_path {continue_config_path} --model_path {continue_restore_path} --out_path {out_wav_path}"
+run_cli(inference_command)
+
 # restore the model and continue training for one more epoch
 command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_tts.py --continue_path {continue_path} "
 run_cli(command_train)

tests/tts_tests/test_tacotron2_d-vectors_train.py

@@ -4,6 +4,7 @@ import shutil
 
 from tests import get_device_id, get_tests_output_path, run_cli
 from TTS.tts.configs.tacotron2_config import Tacotron2Config
+from TTS.utils.trainer_utils import get_last_checkpoint
 
 config_path = os.path.join(get_tests_output_path(), "test_model_config.json")
 output_path = os.path.join(get_tests_output_path(), "train_outputs")
@@ -52,6 +53,16 @@ run_cli(command_train)
 # Find latest folder
 continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
 
+# Inference using TTS API
+continue_config_path = os.path.join(continue_path, "config.json")
+continue_restore_path, _ = get_last_checkpoint(continue_path)
+out_wav_path = os.path.join(get_tests_output_path(), 'output.wav')
+speaker_id = "ljspeech-1"
+continue_speakers_path = config.d_vector_file
+
+inference_command = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' tts --text 'This is an example.' --speaker_idx {speaker_id} --speakers_file_path {continue_speakers_path} --config_path {continue_config_path} --model_path {continue_restore_path} --out_path {out_wav_path}"
+run_cli(inference_command)
+
 # restore the model and continue training for one more epoch
 command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_tts.py --continue_path {continue_path} "
 run_cli(command_train)

tests/tts_tests/test_tacotron2_speaker_emb_train.py

@@ -4,6 +4,7 @@ import shutil
 
 from tests import get_device_id, get_tests_output_path, run_cli
 from TTS.tts.configs.tacotron2_config import Tacotron2Config
+from TTS.utils.trainer_utils import get_last_checkpoint
 
 config_path = os.path.join(get_tests_output_path(), "test_model_config.json")
 output_path = os.path.join(get_tests_output_path(), "train_outputs")
@@ -49,6 +50,16 @@ run_cli(command_train)
 # Find latest folder
 continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
 
+# Inference using TTS API
+continue_config_path = os.path.join(continue_path, "config.json")
+continue_restore_path, _ = get_last_checkpoint(continue_path)
+out_wav_path = os.path.join(get_tests_output_path(), 'output.wav')
+speaker_id = "ljspeech-1"
+continue_speakers_path = os.path.join(continue_path, "speakers.json")
+
+inference_command = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' tts --text 'This is an example.' --speaker_idx {speaker_id} --speakers_file_path {continue_speakers_path} --config_path {continue_config_path} --model_path {continue_restore_path} --out_path {out_wav_path}"
+run_cli(inference_command)
+
 # restore the model and continue training for one more epoch
 command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_tts.py --continue_path {continue_path} "
 run_cli(command_train)

tests/tts_tests/test_tacotron2_train.py

@@ -4,6 +4,7 @@ import shutil
 
 from tests import get_device_id, get_tests_output_path, run_cli
 from TTS.tts.configs.tacotron2_config import Tacotron2Config
+from TTS.utils.trainer_utils import get_last_checkpoint
 
 config_path = os.path.join(get_tests_output_path(), "test_model_config.json")
 output_path = os.path.join(get_tests_output_path(), "train_outputs")
@@ -47,6 +48,14 @@ run_cli(command_train)
 # Find latest folder
 continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
 
+# Inference using TTS API
+continue_config_path = os.path.join(continue_path, "config.json")
+continue_restore_path, _ = get_last_checkpoint(continue_path)
+out_wav_path = os.path.join(get_tests_output_path(), 'output.wav')
+
+inference_command = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' tts --text 'This is an example.' --config_path {continue_config_path} --model_path {continue_restore_path} --out_path {out_wav_path}"
+run_cli(inference_command)
+
 # restore the model and continue training for one more epoch
 command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_tts.py --continue_path {continue_path} "
 run_cli(command_train)

tests/tts_tests/test_tacotron_train.py

@@ -4,6 +4,7 @@ import shutil
 
 from tests import get_device_id, get_tests_output_path, run_cli
 from TTS.tts.configs.tacotron_config import TacotronConfig
+from TTS.utils.trainer_utils import get_last_checkpoint
 
 config_path = os.path.join(get_tests_output_path(), "test_model_config.json")
 output_path = os.path.join(get_tests_output_path(), "train_outputs")
@@ -48,6 +49,14 @@ run_cli(command_train)
 # Find latest folder
 continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
 
+# Inference using TTS API
+continue_config_path = os.path.join(continue_path, "config.json")
+continue_restore_path, _ = get_last_checkpoint(continue_path)
+out_wav_path = os.path.join(get_tests_output_path(), 'output.wav')
+
+inference_command = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' tts --text 'This is an example.' --config_path {continue_config_path} --model_path {continue_restore_path} --out_path {out_wav_path}"
+run_cli(inference_command)
+
 # restore the model and continue training for one more epoch
 command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_tts.py --continue_path {continue_path} "
 run_cli(command_train)

tests/tts_tests/test_vits_multilingual_speaker_emb_train.py

@@ -5,6 +5,7 @@ import shutil
 from tests import get_device_id, get_tests_output_path, run_cli
 from TTS.config.shared_configs import BaseDatasetConfig
 from TTS.tts.configs.vits_config import VitsConfig
+from TTS.utils.trainer_utils import get_last_checkpoint
 
 config_path = os.path.join(get_tests_output_path(), "test_model_config.json")
 output_path = os.path.join(get_tests_output_path(), "train_outputs")
@@ -34,7 +35,7 @@ config = VitsConfig(
     text_cleaner="english_cleaners",
     use_phonemes=True,
     use_espeak_phonemes=True,
-    phoneme_language="en-us",
+    phoneme_language="en",
     phoneme_cache_path="tests/data/ljspeech/phoneme_cache/",
     run_eval=True,
     test_delay_epochs=-1,
@@ -82,6 +83,18 @@ run_cli(command_train)
 # Find latest folder
 continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
 
+# Inference using TTS API
+continue_config_path = os.path.join(continue_path, "config.json")
+continue_restore_path, _ = get_last_checkpoint(continue_path)
+out_wav_path = os.path.join(get_tests_output_path(), 'output.wav')
+speaker_id = "ljspeech"
+languae_id = "en"
+continue_speakers_path = os.path.join(continue_path, "speakers.json")
+continue_languages_path = os.path.join(continue_path, "language_ids.json")
+
+inference_command = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' tts --text 'This is an example.' --speaker_idx {speaker_id} --speakers_file_path {continue_speakers_path} --language_ids_file_path {continue_languages_path} --language_idx {languae_id} --config_path {continue_config_path} --model_path {continue_restore_path} --out_path {out_wav_path}"
+run_cli(inference_command)
+
 # restore the model and continue training for one more epoch
 command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_tts.py --continue_path {continue_path} "
 run_cli(command_train)

tests/tts_tests/test_vits_multilingual_train-d_vectors.py

@@ -5,13 +5,14 @@ import shutil
 from tests import get_device_id, get_tests_output_path, run_cli
 from TTS.config.shared_configs import BaseDatasetConfig
 from TTS.tts.configs.vits_config import VitsConfig
+from TTS.utils.trainer_utils import get_last_checkpoint
 
 config_path = os.path.join(get_tests_output_path(), "test_model_config.json")
 output_path = os.path.join(get_tests_output_path(), "train_outputs")
 
 
 dataset_config_en = BaseDatasetConfig(
-    name="ljspeech",
+    name="ljspeech_test",
     meta_file_train="metadata.csv",
     meta_file_val="metadata.csv",
     path="tests/data/ljspeech",
@@ -19,7 +20,7 @@ dataset_config_en = BaseDatasetConfig(
 )
 
 dataset_config_pt = BaseDatasetConfig(
-    name="ljspeech",
+    name="ljspeech_test",
     meta_file_train="metadata.csv",
     meta_file_val="metadata.csv",
     path="tests/data/ljspeech",
@@ -31,7 +32,7 @@ config = VitsConfig(
     eval_batch_size=2,
     num_loader_workers=0,
     num_eval_loader_workers=0,
-    text_cleaner="english_cleaners",
+    text_cleaner="multilingual_cleaners",
     use_phonemes=False,
     phoneme_cache_path="tests/data/ljspeech/phoneme_cache/",
     run_eval=True,
@@ -85,6 +86,18 @@ run_cli(command_train)
 # Find latest folder
 continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
 
+# Inference using TTS API
+continue_config_path = os.path.join(continue_path, "config.json")
+continue_restore_path, _ = get_last_checkpoint(continue_path)
+out_wav_path = os.path.join(get_tests_output_path(), 'output.wav')
+speaker_id = "ljspeech-1"
+languae_id = "en"
+continue_speakers_path = config.d_vector_file
+continue_languages_path = os.path.join(continue_path, "language_ids.json")
+
+inference_command = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' tts --text 'This is an example.' --speaker_idx {speaker_id} --speakers_file_path {continue_speakers_path} --language_ids_file_path {continue_languages_path} --language_idx {languae_id} --config_path {continue_config_path} --model_path {continue_restore_path} --out_path {out_wav_path}"
+run_cli(inference_command)
+
 # restore the model and continue training for one more epoch
 command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_tts.py --continue_path {continue_path} "
 run_cli(command_train)

tests/tts_tests/test_vits_speaker_emb_train.py

@@ -4,6 +4,7 @@ import shutil
 
 from tests import get_device_id, get_tests_output_path, run_cli
 from TTS.tts.configs.vits_config import VitsConfig
+from TTS.utils.trainer_utils import get_last_checkpoint
 
 config_path = os.path.join(get_tests_output_path(), "test_model_config.json")
 output_path = os.path.join(get_tests_output_path(), "train_outputs")
@@ -25,7 +26,7 @@ config = VitsConfig(
     print_step=1,
     print_eval=True,
     test_sentences=[
-        ["Be a voice, not an echo.", "ljspeech"],
+        ["Be a voice, not an echo.", "ljspeech-1"],
     ],
 )
 # set audio config
@@ -45,7 +46,7 @@ config.save_json(config_path)
 command_train = (
     f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_tts.py --config_path {config_path} "
     f"--coqpit.output_path {output_path} "
-    "--coqpit.datasets.0.name ljspeech "
+    "--coqpit.datasets.0.name ljspeech_test "
     "--coqpit.datasets.0.meta_file_train metadata.csv "
     "--coqpit.datasets.0.meta_file_val metadata.csv "
     "--coqpit.datasets.0.path tests/data/ljspeech "
@@ -57,6 +58,16 @@ run_cli(command_train)
 # Find latest folder
 continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
 
+# Inference using TTS API
+continue_config_path = os.path.join(continue_path, "config.json")
+continue_restore_path, _ = get_last_checkpoint(continue_path)
+out_wav_path = os.path.join(get_tests_output_path(), 'output.wav')
+speaker_id = "ljspeech-1"
+continue_speakers_path = os.path.join(continue_path, "speakers.json")
+
+inference_command = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' tts --text 'This is an example.' --speaker_idx {speaker_id} --speakers_file_path {continue_speakers_path} --config_path {continue_config_path} --model_path {continue_restore_path} --out_path {out_wav_path}"
+run_cli(inference_command)
+
 # restore the model and continue training for one more epoch
 command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_tts.py --continue_path {continue_path} "
 run_cli(command_train)

tests/tts_tests/test_vits_train.py

@@ -4,6 +4,7 @@ import shutil
 
 from tests import get_device_id, get_tests_output_path, run_cli
 from TTS.tts.configs.vits_config import VitsConfig
+from TTS.utils.trainer_utils import get_last_checkpoint
 
 config_path = os.path.join(get_tests_output_path(), "test_model_config.json")
 output_path = os.path.join(get_tests_output_path(), "train_outputs")
@@ -48,6 +49,14 @@ run_cli(command_train)
 # Find latest folder
 continue_path = max(glob.glob(os.path.join(output_path, "*/")), key=os.path.getmtime)
 
+# Inference using TTS API
+continue_config_path = os.path.join(continue_path, "config.json")
+continue_restore_path, _ = get_last_checkpoint(continue_path)
+out_wav_path = os.path.join(get_tests_output_path(), 'output.wav')
+
+inference_command = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' tts --text 'This is an example.' --config_path {continue_config_path} --model_path {continue_restore_path} --out_path {out_wav_path}"
+run_cli(inference_command)
+
 # restore the model and continue training for one more epoch
 command_train = f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_tts.py --continue_path {continue_path} "
 run_cli(command_train)