CHANGELOG.md

Changelog

v3.10.3 (2023-04-19)

Fix

• train: Don't save model when tuning for auto batch size (2311a35)

v3.10.2 (2023-04-19)

Fix

• train: Properly stop training after epochs has been reached (f9bb3d8)

v3.10.1 (2023-04-19)

Fix

• train: Support ddp in Windows (gloo backend) (#394) (bcb0507)

v3.10.0 (2023-04-18)

Feature

• Replace fairseq with transformers (#385) (a2fe0f3)

v3.9.5 (2023-04-18)

Fix

• train: Set persistent_workers = True in DataLoader for performance, do not save checkpoints, fix logging issue and multiple warning issues, do not do validation when global_step == 0 (#384) (6cab9af)

v3.9.4 (2023-04-18)

Fix

• gui: Always use "spawn" context in ProcessPool (#383) (5d7fb77)

v3.9.3 (2023-04-16)

Fix

• gui: Fix subprocess errors in linux and fix wrong error logging (#358) (fd67db6)

v3.9.2 (2023-04-16)

Fix

• train: Fix y_mel length (#356) (2d71992)

v3.9.1 (2023-04-16)

Fix

• train: Allow higher segment size (#351) (09d5a52)
• cluster: Do not use weights_only in get_cluster_model() (#354) (24c05d1)

v3.9.0 (2023-04-16)

Feature

• train: Add option to name ckpts by epochs (#353) (bba24c4)

v3.8.1 (2023-04-16)

Fix

• Patch stft and add mps to get_optimal_device() (#352) (da928aa)

v3.8.0 (2023-04-15)

Feature

• train: Automatically decide batch_size (#342) (8ffa128)

v3.7.3 (2023-04-15)

Fix

• gui: Show errors raised in inference (#341) (99833c5)

v3.7.2 (2023-04-15)

Fix

• train: Suppress pytorch logs for deprecated TypedStorage (#336) (e67ac62)

v3.7.1 (2023-04-15)

Fix

• train: Fix check for notebook / Colab (#335) (7f69814)

v3.7.0 (2023-04-14)

Feature

• train: Add option to specify TensorBoardLogger version parameter support (#331) (a685123)

v3.6.2 (2023-04-14)

Fix

• Fix torch.load and save to use file objects and weights_only and remove unidecode (#327) (4aad701)

v3.6.1 (2023-04-14)

Fix

• train: Fix gradient logging (#320) (73ef3dc)

v3.6.0 (2023-04-13)

Feature

• realtime: Support sola algorithm (#316) (0fcbf99)

v3.5.1 (2023-04-13)

Fix

• logger: Do not use rich in notebook (#317) (03c8240)

v3.5.0 (2023-04-13)

Feature

• gui: Run inference in thread and disable button (#279) (c55caa8)

v3.4.0 (2023-04-13)

Feature

• train: Make num_workers configurable (#315) (e8df714)

v3.3.1 (2023-04-13)

Performance

• train: Specify num_workers in DataLoader (#313) (6042164)

v3.3.0 (2023-04-13)

Feature

• train: Use RichProgressBar (#311) (17e937a)

v3.2.0 (2023-04-13)

Feature

• train: Add optional accumulate_grad_batches config param (#306) (1172b23)

v3.1.13 (2023-04-12)

Fix

• logger: Fix too noisy logger (#300) (bd0eb33)
• main: Fix cli() not called in main (#299) (11f2d24)

v3.1.12 (2023-04-12)

Fix

• train: Fix ddp not working (#297) (bec43fc)

v3.1.11 (2023-04-12)

Fix

• logger: Fix init_logger not showing debug messages in certain conditions as intended (#296) (d3ab7d6)

v3.1.10 (2023-04-11)

Fix

• notebooks: Improves inference (#286) (d3228df)
• train: Step lr schedulers at end of epoch (#291) (3af223e)

v3.1.9 (2023-04-10)

Fix

• train: Fix fp16_run not being mix precision and fix bf16 errors (#283) (b0dd0ed)

v3.1.8 (2023-04-10)

Fix

• readme: Fix wrong commands in "Before training" (#280) (e056ad9)

v3.1.7 (2023-04-09)

Fix

• train: Improve quality of training (#274) (7ed71d6)

v3.1.6 (2023-04-09)

Fix

• train: Fix checkpoint not properly loaded (#271) (0979147)

v3.1.5 (2023-04-09)

Fix

• train: Fix optim_d functions called in wrong order (#269) (13d6346)

v3.1.4 (2023-04-09)

Fix

• train: Add bf16 and fp16 support (#268) (4229fd8)

v3.1.3 (2023-04-09)

Fix

• deps: Update dependency starlette to v0.26.1 (#256) (5eb574b)

v3.1.2 (2023-04-09)

Fix

• Remove wrong test and trigger release (#266) (9ea77e4)
• deps: Remove pydantic constraints (#263) (f446e3b)
• Fix fastapi version to 0.88 (#261) (a26f387)
• utils: Fix get_optimal_device (#262) (79e4b5a)

v3.1.1 (2023-04-08)

Fix

• deps: Update dependency fastapi to <0.96 (#255) (29c8cc0)

v3.1.0 (2023-04-08)

Feature

• Migrate to lightning (#246) (824ecbd)

v3.0.5 (2023-04-08)

Fix

• cluster: Fix train_cluster (#250) (b0c93e4)

v3.0.4 (2023-04-06)

Fix

• main: Fix default config type to revert breaking changes (#243) (e05c0b5)

v3.0.3 (2023-04-05)

Fix

• Fix issues when loading legacy checkpoint and fix pre-hubert n_jobs (#236) (15f1e7f)

v3.0.2 (2023-04-04)

Performance

• Move methods from dataloader to pre-hubert (#218) (d5a4456)

v3.0.1 (2023-04-03)

Fix

• Remove possible leak in unused code (#203) (e921c3d)

Performance

• utils: Better implementation of repeat_expand_2d (#216) (ef30a9d)

v3.0.0 (2023-04-03)

Feature

• Add quickvc, fix usage of ContentVec, remove onnx support (#197) (1a6c021)

Breaking

• The module structure has been completely changed, and some of the default values for configurations have also been changed. (1a6c021)

Documentation

• readme: Update allcontributors link for @MashiroSA (#210) (650524b)
• readme: Add paperspace referral (#209) (7280012)
• readme: Add paperspace link and add more description, add a link for zh-CN docs (#208) (bc4b122)

v2.1.5 (2023-04-01)

Fix

• deps: Update dependency tensorboard to v2.12.1 (#199) (0ccda1c)

v2.1.4 (2023-03-31)

Fix

• deps: Update dependency gradio to v3.24.1 (#196) (4fa141b)

v2.1.3 (2023-03-31)

Fix

• deps: Update dependency gradio to v3.24.0 (#192) (4e441cb)

Documentation

• Add sbersier as a contributor for ideas, and userTesting (#185) (a655bf4)
• Add ColdCawfee as a contributor for bug (#177) (87a09e6)

v2.1.2 (2023-03-28)

Fix

• gui: Fix wrong devices set as default (#176) (6265f8f)
• main: Fix -h option overridden (#175) (52f1cfe)

Documentation

• Add xieyumc as a contributor for doc (#174) (29474d9)
• readme_zh_cn: Update README_zh_CN.md (#173) (f94a14c)

v2.1.1 (2023-03-27)

Fix

• deps: Update dependency rich to v13.3.3 (#171) (8bdefa9)

Documentation

• Add nerdyrodent as a contributor for video (#168) (78ab661)
• Add heyfixit as a contributor for doc (#167) (32a2a63)
• Add Desuka-art as a contributor for bug (#165) (fe3c6bf)
• Add ruckusmattster as a contributor for bug (#163) (2b971db)
• Add pierluigizagaria as a contributor for userTesting (#160) (6fabe8d)
• Add Satisfy256 as a contributor for bug (#158) (ee72aee)
• Add DL909 as a contributor for bug (#152) (a5e6651)

v2.1.0 (2023-03-27)

Feature

• main: Add an option to launch tensorboard in train command (#149) (ef22cce)

v2.0.0 (2023-03-27)

Fix

• Fix preprocessing and convert bool options to flags, use unidecode to decode non-ASCII filenames in pre-resample (#147) (98d7ee2)

Breaking

• Preprocessing default parameters changed and some options became flags. (98d7ee2)

Documentation

• readme: Add yt tutorial vid link (#143) (1694f44)

v1.4.3 (2023-03-26)

Performance

• preprocess: Specify samplerate to reduce memory usage (#137) (6217eda)

v1.4.2 (2023-03-26)

Fix

• logging: Initialize logging in logger file and move version log (#131) (441d51f)
• preprocess: Fix dtype in sf.read() to save memory and fix preprocess_resample (#132) (0af1e13)
• Fix audio resampled to 22kHz (#130) (4203f37)

Documentation

• Update notebook and README.md (#128) (38d9744)

v1.4.1 (2023-03-26)

Fix

• main,gui: Fix some parameters not passed (#124) (6cfe3d3)

v1.4.0 (2023-03-26)

Feature

• preprocessing: Add 2 more preprocessing commands (#123) (45eba0f)

Documentation

• Add BlueAmulet as a contributor for code (#119) (6a7e8ba)

v1.3.5 (2023-03-26)

Fix

• data_utils: Allow float32 audio to be processed properly (#115) (13943b6)

v1.3.4 (2023-03-25)

Fix

• Change default f0 method from crepe to dio (#100) (baf58d2)

v1.3.3 (2023-03-25)

Fix

• Fix old checkpoint deletion by sorting the models properly (#65) (#89) (287dc94)

Documentation

• Add Lordmau5 as a contributor for bug, and code (#102) (4df46ee)

v1.3.2 (2023-03-24)

Fix

• gui: Fix devices list and fix tqdm error in gui (#99) (59724cd)

Documentation

• Add MashiroSA as a contributor for doc, and bug (#98) (495b7cb)
• notebook: Fix cluster inference command and improve cluster training command (#93) (7642594)

v1.3.1 (2023-03-24)

Fix

• gui: Fix defaut for auto_play (#97) (07920a4)
• infer_tool: Fix speaker not automaticlly set to the first one if not found in cluster inference (#96) (a643e4f)

Documentation

• notebook: Add cluster training and inference (#84) (9ffb621)

v1.3.0 (2023-03-23)

Feature

• gui: Better error handling (#85) (985704b)

v1.2.11 (2023-03-23)

Fix

• gui,onnx: Fix onnx export and fix gui (#83) (3e9a47d)

v1.2.10 (2023-03-23)

Fix

• cluster: Fix cluster not working (#82) (29b209c)

v1.2.9 (2023-03-23)

Fix

• gui: Fix speakers and devices not updated and fix default presets (#80) (a851150)

v1.2.8 (2023-03-22)

Fix

• deps: Update dependency torchcrepe to v0.0.18 (#76) (4fda479)

Documentation

• readme: Shorten docs (#75) (e0c1572)

v1.2.7 (2023-03-22)

Fix

• Fix clean_checkpoints (#66) (e5169bf)

v1.2.6 (2023-03-22)

Fix

• gui: Disable checkbox if CUDA is not available and show errors for VC (#71) (3fdd983)

Documentation

• Add BlueAmulet as a contributor for question (#74) (8d073e3)
• Add GarrettConway as a contributor for doc (#73) (6c6cbc6)
• readme: Update README.md regarding installation, update. WSL audio support (#72) (4f1323b)

v1.2.5 (2023-03-22)

Fix

• Fix RTF calculation (#70) (fb25500)

v1.2.4 (2023-03-22)

Fix

• utils: Fix latest_checkpoint_path (#64) (00b9f4a)

v1.2.3 (2023-03-21)

Fix

• deps: Update dependency onnxsim to v0.4.19 (#62) (f8a4cf6)

v1.2.2 (2023-03-21)

Fix

• deps: Update dependency onnxoptimizer to v0.3.10 (#61) (d0137f9)

Documentation

• readme: Add links for pretrained models and fix GUI pic height (#60) (34ac39f)
• notebook: Add more explanation to notebook (#59) (9b3c483)

v1.2.1 (2023-03-21)

Fix

• Use librosa.load() instead of soundfile.read() (#58) (b343106)
• gui: Fix window too big to show in a FHD environment (#57) (259e6e6)

v1.2.0 (2023-03-21)

Feature

• Add presets (#55) (e8adcc6)

v1.1.1 (2023-03-21)

Fix

• deps: Update dependency gradio to v3.23.0 (#54) (a2bdb48)

v1.1.0 (2023-03-21)

Feature

• Enhance RealtimeVC (#52) (81551ce)

Documentation

• Update gui screenshot (#53) (58d06aa)

v1.0.2 (2023-03-21)

Fix

• deps: Update dependency scipy to v1.10.1 (#35) (e0253bf)

v1.0.1 (2023-03-20)

Documentation

• Add ThrowawayAccount01 as a contributor for bug (#47) (15e31fa)
• Add BlueAmulet as a contributor for ideas (#46) (a3bcb2b)

Performance

• inference_main: Do dummy inference before running vc (#45) (4066c43)

v1.0.0 (2023-03-20)

Fix

• Fix default dataset path (#43) (ac47fed)

Breaking

• the behaviour of preprocess_resample changes when there is a folder ./dataset_raw/44k and "44k" is no longer allowed as a speaker name in some conditions (ac47fed)

v0.8.2 (2023-03-20)

Fix

• Fix compute_f0_crepe returning wrong length (#42) (afb42b0)

v0.8.1 (2023-03-20)

Fix

• deps: Update dependency librosa to v0.10.0 (#40) (8e92f71)

v0.8.0 (2023-03-20)

Feature

• Add more f0 calculation methods (#39) (6b3b20d)

v0.7.1 (2023-03-20)

Fix

• deps: Update dependency gradio to v3.22.1 (#33) (f09fc23)

v0.7.0 (2023-03-20)

Feature

• preprocessing: Allow nested dataset (#19) (0433151)

v0.6.3 (2023-03-20)

Fix

• deps: Update dependency torch to v1.13.1 (#27) (8826d68)
• deps: Update dependency torchaudio to v0.13.1 (#28) (989f5d9)

Documentation

• readme: Update notes about VRAM caps (#18) (0a245f4)

v0.6.2 (2023-03-19)

Fix

• Use hubert preprocess force_rebuild argument (#15) (87cf807)

Documentation

• Add GarrettConway as a contributor for bug (#17) (31d9671)
• notebook: Launch tensorboard (#16) (52229ba)
• Add 34j as a contributor for example, infra, and 6 more (#14) (1b90378)
• Add GarrettConway as a contributor for code (#13) (716813f)

v0.6.1 (2023-03-19)

Performance

• preprocessing: Better performance (#12) (668c8e1)

v0.6.0 (2023-03-18)

Feature

• Configurable input and output devices (#11) (a822a60)

Documentation

• notebook: Fix notebook (#9) (427b4c1)
• Update notebook (#8) (ae3e471)

v0.5.0 (2023-03-18)

Feature

• gui: Remember last directory (misc) (92558da)
• main: Show defaults (3d298df)

Fix

• main: Fix option names (7ff34fe)

Documentation

• readme: Update README.md (#6) (b988101)

v0.4.1 (2023-03-18)

Fix

• Call init_logger() (#5) (e6378f1)

v0.4.0 (2023-03-18)

Feature

• Enhance realtime algorythm (#4) (d789a12)

v0.3.0 (2023-03-17)

Feature

• Add gui (#3) (34aec2b)

Documentation

• Update notebook (7b74606)

v0.2.1 (2023-03-17)

Fix

• notebook: Fix notebook (3ed00cc)

v0.2.0 (2023-03-17)

Feature

• Realtime inference (#2) (4dea1ae)

v0.1.0 (2023-03-17)

Feature

• Main feat (#1) (faa990c)