Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

XTTS v1.1 GPT Trainer #3086

Merged
merged 24 commits into from
Oct 25, 2023
Merged
Changes from 1 commit
Commits
Show all changes
24 commits
Select commit Hold shift + click to select a range
a32961b
Add XTTS base training code
Edresson Oct 11, 2023
40a4e63
Update mel spectrogram for the style encoder
Edresson Oct 11, 2023
47d613d
Add reproducible evaluation
Edresson Oct 13, 2023
bafab04
Add prompting masking
Edresson Oct 16, 2023
2f868dd
Bug fix on reproducible evaluation
Edresson Oct 16, 2023
c4ceaab
Add test sentences during the training
Edresson Oct 16, 2023
9e3598c
Bug Fix on inference using XTTS trainer checkpoint
Edresson Oct 18, 2023
469d624
Update LJspeech XTTS recipe
Edresson Oct 18, 2023
5f98dbe
Update Ljspeech XTTS recipe
Edresson Oct 18, 2023
94dcf84
Rename XTTS recipe
Edresson Oct 18, 2023
1f92741
Fix issue #2971
Edresson Oct 18, 2023
affaf11
Add XTTS training unit test
Edresson Oct 18, 2023
ec7f547
Rebase bug fix and update recipe
Edresson Oct 21, 2023
e8a1a50
Remove unused vars in Delightful TTS layers tests
Edresson Oct 23, 2023
653f2e7
Update xtts trainer recipe
Edresson Oct 23, 2023
8853e1c
Update XTTS recipe to only download checkpoint if it is needed
Edresson Oct 23, 2023
6fefc36
Update XTTS docs
Edresson Oct 23, 2023
1ee8096
Update XTTS docs
Edresson Oct 23, 2023
37b7945
Update XTTS train not implemented error to point to the XTTS docs
Edresson Oct 23, 2023
67ca70a
Fix Delightful TTS layers unit test
Edresson Oct 23, 2023
0f96abb
Add FT inference example on XTTS docs
Edresson Oct 23, 2023
de1d521
Update XTTS docs
Edresson Oct 23, 2023
8af3d2d
Add a dedicated workflow for XTTS tests
Edresson Oct 24, 2023
01839af
Bug fix on XTTS masking training
Edresson Oct 24, 2023
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Prev Previous commit
Next Next commit
Update XTTS recipe to only download checkpoint if it is needed
  • Loading branch information
Edresson committed Oct 23, 2023
commit 8853e1c3ecce7feb1382069800bd1f16027a7921
19 changes: 13 additions & 6 deletions recipes/ljspeech/xtts_v1/train_gpt_xtts.py
Original file line number Diff line number Diff line change
Expand Up @@ -44,24 +44,31 @@
# DVAE files
DVAE_CHECKPOINT_LINK = "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v1/v1.1.1/dvae.pth"
MEL_NORM_LINK = "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v1/v1.1.1/mel_stats.pth"
# download DVAE files
print(" > Downloading DVAE files!")
ModelManager._download_model_files([MEL_NORM_LINK, DVAE_CHECKPOINT_LINK], CHECKPOINTS_OUT_PATH, progress_bar=True)

# Set the path to the downloaded files
DVAE_CHECKPOINT = os.path.join(CHECKPOINTS_OUT_PATH, DVAE_CHECKPOINT_LINK.split("/")[-1])
MEL_NORM_FILE = os.path.join(CHECKPOINTS_OUT_PATH, MEL_NORM_LINK.split("/")[-1])

# Download XTTS v1.1 checkpoint
# download DVAE files if needed
if not os.path.isfile(DVAE_CHECKPOINT) or not os.path.isfile(MEL_NORM_FILE):
print(" > Downloading DVAE files!")
ModelManager._download_model_files([MEL_NORM_LINK, DVAE_CHECKPOINT_LINK], CHECKPOINTS_OUT_PATH, progress_bar=True)


# Download XTTS v1.1 checkpoint if needed
TOKENIZER_FILE_LINK = "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v1/v1.1.1/vocab.json"
XTTS_CHECKPOINT_LINK = "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v1/v1.1.1/model.pth"
print(" > Downloading XTTS v1.1 files!")
ModelManager._download_model_files([TOKENIZER_FILE_LINK, XTTS_CHECKPOINT_LINK], CHECKPOINTS_OUT_PATH, progress_bar=True)

# XTTS transfer learning parameters: You we need to provide the paths of XTTS model checkpoint that you want to do the fine tuning.
TOKENIZER_FILE = os.path.join(CHECKPOINTS_OUT_PATH, TOKENIZER_FILE_LINK.split("/")[-1]) # vocab.json file
XTTS_CHECKPOINT = os.path.join(CHECKPOINTS_OUT_PATH, XTTS_CHECKPOINT_LINK.split("/")[-1]) # model.pth file

# download XTTS v1.1 files if needed
if not os.path.isfile(TOKENIZER_FILE) or not os.path.isfile(XTTS_CHECKPOINT):
print(" > Downloading XTTS v1.1 files!")
ModelManager._download_model_files([TOKENIZER_FILE_LINK, XTTS_CHECKPOINT_LINK], CHECKPOINTS_OUT_PATH, progress_bar=True)


# Training sentences generations
SPEAKER_REFERENCE = (
"./tests/data/ljspeech/wavs/LJ001-0002.wav" # speaker reference to be used in training test sentences
Expand Down
Loading