storytold/storyteller-ml
2
0
Fork 0
mirror of https://github.com/storytold/storyteller-ml.git synced 2026-10-28 09:08:54 +00:00
Code Issues Packages Projects Releases Wiki Activity

Deleted comments to reduce confusion.

Browse Source 
renamed model to encodec_pytorch_model
This commit is contained in:
Michael Chung
2023-10-18 00:16:50 -04:00
parent 7de861c93e
commit e685ace23a
2 changed files with 8 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tts/VALL-E-X/main.py
+7 -13
View File
@@ -1,11 +1,5 @@
import argparse import argparse
# python3 main.py -create_prompt_with_list 20.wav 21.wav -create_prompt_name alice
# python3 main.py -use_prompt_name alice -output_file_name alice_says.wav -text "This is the text to be converted to speech."
# python3 main.py -create_prompt_with_list 20.wav 21.wav 20.wav 21.wav -create_prompt_name alice -use_prompt_name alice -output_file_name alice_says.wav -text "This is my voice."
# python3 main.py -create_prompt_with_list goku.wav -create_prompt_name goku -use_prompt_name goku -output_file_name goku_test.wav -text "oh my god krillin! where did you put my sensu beans!?"
# python3 main.py -create_prompt_with_list
from typing import List from typing import List
import pathlib import pathlib
@@ -106,7 +100,6 @@ class VoiceDesigner():
# multiple audio files concat # multiple audio files concat
# concat audio and trim the audio if it is longer than 14-15 seconds# if prompt is too long it will error out anyways # concat audio and trim the audio if it is longer than 14-15 seconds# if prompt is too long it will error out anyways
if len(audio_file_paths) > 0: if len(audio_file_paths) > 0:
# TODO: going to assume brandon renames the files as they are uploaded
# try:# ./customs/ <-- folder it adds by default until we modified. # try:# ./customs/ <-- folder it adds by default until we modified.
# load the audio files and then concat them. # load the audio files and then concat them.
tensor_result = AudioOps.stitch_wav_files_resample(files=audio_file_paths) tensor_result = AudioOps.stitch_wav_files_resample(files=audio_file_paths)
@@ -220,23 +213,24 @@ if __name__ == "__main__":
main(args) main(args)
# /home/tensor/code/storyteller/storyteller-ml/tts/VALL-E-X/.venv/bin/python # inference example
# /home/tensor/code/storyteller/storyteller-ml/tts/VALL-E-X/main.py #/home/tensor/code/storyteller/storyteller-ml/tts/VALL-E-X/.venv/bin/python /home/tensor/code/storyteller/storyteller-ml/tts/VALL-E-X/main.py --text "hello world" --mode 0 --whisper-folder-path "/home/tensor/code/storyteller/storyteller-ml/tts/VALL-E-X/whisper" --whisper-model medium --vocos-folder-path "/home/tensor/code/TTSDockerContainer/Vall-E-mount/models/vocos-encodec-24khz" --vallex-path "/home/tensor/code/TTSDockerContainer/Vall-E-mount/models/vallex-checkpoint.pt" --prompt-path "/home/tensor/code/storyteller/VALL-E-X-TTS-Container/Vall-E-mount/prompts" --prompt-name "goku.npz" --audio-path "/home/tensor/code/storyteller/VALL-E-X-TTS-Container/Vall-E-mount/output/" --audio-name "hello_world_test.wav" --tmp-work-dir "/tmp"
#/home/tensor/code/storyteller/storyteller-ml/tts/VALL-E-X/.venv/bin/python
#/home/tensor/code/storyteller/storyteller-ml/tts/VALL-E-X/main.py
# --text "hello world" # --text "hello world"
# --mode 0 # --mode 0
# --audio-wav-files "/home/tensor/code/TTSDockerContainer/Vall-E-mount/input/20.wav" "/home/tensor/code/TTSDockerContainer/Vall-E-mount/input/21.wav"
# --whisper-folder-path "/home/tensor/code/storyteller/storyteller-ml/tts/VALL-E-X/whisper" # --whisper-folder-path "/home/tensor/code/storyteller/storyteller-ml/tts/VALL-E-X/whisper"
# --whisper-model medium # --whisper-model medium
# --vocos-folder-path "/home/tensor/code/TTSDockerContainer/Vall-E-mount/models/vocos-encodec-24khz" # --vocos-folder-path "/home/tensor/code/TTSDockerContainer/Vall-E-mount/models/vocos-encodec-24khz"
# --vallex-path "/home/tensor/code/TTSDockerContainer/Vall-E-mount/models/vallex-checkpoint.pt" # --vallex-path "/home/tensor/code/TTSDockerContainer/Vall-E-mount/models/vallex-checkpoint.pt"
# --prompt-path "/home/tensor/code/storyteller/VALL-E-X-TTS-Container/Vall-E-mount/prompts" # --prompt-path "/home/tensor/code/storyteller/VALL-E-X-TTS-Container/Vall-E-mount/prompts"
# --prompt-name "goku.npz" # --prompt-name "goku"
# --audio-path "/home/tensor/code/storyteller/VALL-E-X-TTS-Container/Vall-E-mount/output/" # --audio-path "/home/tensor/code/storyteller/VALL-E-X-TTS-Container/Vall-E-mount/output/"
# --audio-name "hello_world_test.wav" # --audio-name "hello_world_test.wav"
# --tmp-work-dir "/tmp" # --tmp-work-dir "/tmp"
# inference example
#/home/tensor/code/storyteller/storyteller-ml/tts/VALL-E-X/.venv/bin/python /home/tensor/code/storyteller/storyteller-ml/tts/VALL-E-X/main.py --text "hello world" --mode 0 --whisper-folder-path "/home/tensor/code/storyteller/storyteller-ml/tts/VALL-E-X/whisper" --whisper-model medium --vocos-folder-path "/home/tensor/code/TTSDockerContainer/Vall-E-mount/models/vocos-encodec-24khz" --vallex-path "/home/tensor/code/TTSDockerContainer/Vall-E-mount/models/vallex-checkpoint.pt" --prompt-path "/home/tensor/code/storyteller/VALL-E-X-TTS-Container/Vall-E-mount/prompts" --prompt-name "goku.npz" --audio-path "/home/tensor/code/storyteller/VALL-E-X-TTS-Container/Vall-E-mount/output/" --audio-name "hello_world_test.wav" --tmp-work-dir "/tmp"
# create voice example # create voice example
#/home/tensor/code/storyteller/storyteller-ml/tts/VALL-E-X/.venv/bin/python /home/tensor/code/storyteller/storyteller-ml/tts/VALL-E-X/main.py --mode 1 --audio-wav-files "/home/tensor/code/TTSDockerContainer/Vall-E-mount/input/20.wav" "/home/tensor/code/TTSDockerContainer/Vall-E-mount/input/21.wav" --whisper-folder-path "/home/tensor/code/storyteller/storyteller-ml/tts/VALL-E-X/whisper" --whisper-model medium --vocos-folder-path "/home/tensor/code/TTSDockerContainer/Vall-E-mount/models/vocos-encodec-24khz" --vallex-path "/home/tensor/code/TTSDockerContainer/Vall-E-mount/models/vallex-checkpoint.pt" --prompt-path "/home/tensor/code/storyteller/VALL-E-X-TTS-Container/Vall-E-mount/prompts" --prompt-name "test_prompt" --tmp-work-dir "/tmp" #/home/tensor/code/storyteller/storyteller-ml/tts/VALL-E-X/.venv/bin/python /home/tensor/code/storyteller/storyteller-ml/tts/VALL-E-X/main.py --mode 1 --audio-wav-files "/home/tensor/code/TTSDockerContainer/Vall-E-mount/input/20.wav" "/home/tensor/code/TTSDockerContainer/Vall-E-mount/input/21.wav" --whisper-folder-path "/home/tensor/code/storyteller/storyteller-ml/tts/VALL-E-X/whisper" --whisper-model medium --vocos-folder-path "/home/tensor/code/TTSDockerContainer/Vall-E-mount/models/vocos-encodec-24khz" --vallex-path "/home/tensor/code/TTSDockerContainer/Vall-E-mount/models/vallex-checkpoint.pt" --prompt-path "/home/tensor/code/storyteller/VALL-E-X-TTS-Container/Vall-E-mount/prompts" --prompt-name "test_prompt" --tmp-work-dir "/tmp"
tts/VALL-E-X/utils/generation.py
+1 -1
View File
@@ -76,7 +76,7 @@ def load_models_from_pathes(vocos_folder_path:pathlib.Path,vall_e_x_file_path:pa
vocos = None vocos = None
vocos_model = Vocos.from_hparams(vocos_folder_path / pathlib.Path("config.yaml")) vocos_model = Vocos.from_hparams(vocos_folder_path / pathlib.Path("config.yaml"))
state_dict = torch.load(vocos_folder_path / pathlib.Path("pytorch_model.bin"), map_location="cpu") state_dict = torch.load(vocos_folder_path / pathlib.Path("encodec_pytorch_model.bin"), map_location="cpu")
if isinstance(vocos_model.feature_extractor, EncodecFeatures): if isinstance(vocos_model.feature_extractor, EncodecFeatures):
encodec_parameters = { encodec_parameters = {
"feature_extractor.encodec." + key: value "feature_extractor.encodec." + key: value