|
|
4abd6564d1
|
fixed training stats not loading from exported weights, a bit of a readme cleanup, updated example training yaml
|
2023-09-23 19:59:00 -05:00 |
|
|
|
9384900ce6
|
revert the frankensteined "train one model but hotload the other" since it kept loading the last exported weights and I'm not supporting this usecase anymore anyways
|
2023-09-22 13:04:17 -05:00 |
|
|
|
e7da1eb90d
|
edge case
|
2023-09-20 19:20:17 -05:00 |
|
|
|
c0b25541e3
|
restructured some things with the model to remove dead weights
|
2023-09-20 19:10:59 -05:00 |
|
|
|
a6bfe43590
|
added mirostat sampling (given a partially trained model, it got far decent output than I expected, need to test on a better trained model)
|
2023-09-18 18:55:41 -05:00 |
|
|
|
2567e082b5
|
UGH
|
2023-09-16 00:26:13 -05:00 |
|
|
|
22ffaf3a33
|
have loss for the NAR not-ignore the text prompt, I imagine this should help the NAR and explain why it's always had a bit of an issue with training
|
2023-09-15 19:08:44 -05:00 |
|
|
|
4aef798135
|
added picking final candidate based on sum of score instead of first candidate (this changes nothing).
|
2023-09-13 13:19:11 -05:00 |
|
|
|
23a5fdd645
|
implemented a naive beam search (I really should be taking a break)
|
2023-09-12 21:28:07 -05:00 |
|
|
|
a6ae344e5b
|
some comments
|
2023-09-12 16:04:45 -05:00 |
|
|
|
d07c63b9d8
|
unified more things with training the AR+NAR monolothic model
|
2023-09-12 15:54:41 -05:00 |
|
|
|
40ef34e1ca
|
this embedding class definitely works, and migrating from the previous embedding weights seems to work.
|
2023-09-11 14:13:42 -05:00 |
|
|
|
a1f250ffac
|
set default max_levels for NAR to 0 and implicitly set it to max resps levels because the previous way was implicitly assuming all models were outputting at 1+7 RVQ bins.
|
2023-09-10 20:33:33 -05:00 |
|
|
|
671dca88ee
|
throw error when no reference audio is provided in the web UI because someone keeps doing that in the HF space
|
2023-09-10 15:50:50 -05:00 |
|
|
|
ba71020318
|
added option to limit (or exceed) inferenced RVQ-bin levels through the NAR
|
2023-09-10 13:50:13 -05:00 |
|
|
|
c74fe2f718
|
tweaks to web UI
|
2023-09-09 22:27:20 -05:00 |
|
|
|
7f8bd2b936
|
added printing elasped inference time
|
2023-09-09 20:05:03 -05:00 |
|
|
|
4f61f5c889
|
added option to set the trim length for an input prompt
|
2023-09-09 18:04:44 -05:00 |
|
|
|
d10053d11f
|
render README.md markdown for huggingface space
|
2023-09-09 17:04:51 -05:00 |
|
|
|
bc30026377
|
added advanced sampler parameters to the web UI
|
2023-09-09 16:51:36 -05:00 |
|
|
|
5ac119a6e7
|
added light web UI (need to port the telemetry disabling bandaids from aivc)
|
2023-09-09 16:17:20 -05:00 |
|
|
|
10c34c5b98
|
added a length-based decay factor for repetition penalty
|
2023-09-08 21:02:00 -05:00 |
|
|
|
b922f35b6b
|
added documentation on how these new sampling parameters are very iffy and you really need to know what you are doing to use them because this is audio generation and not text generation
|
2023-09-08 20:43:36 -05:00 |
|
|
|
14c78bae39
|
added lots of sampling options (top-k/top-p, repetition penalty, length penalty)
|
2023-09-08 20:30:54 -05:00 |
|
|
|
f69aad9c65
|
some day I'll get it right
|
2023-09-08 15:36:26 -05:00 |
|
|
|
b2907ae7e0
|
seems that my PromEmbedding/RespEmbedding doesn't actually work all that well, naively using dedicated MultiEmbeddings for AR/NAR in the monolithic model is the best way to go
|
2023-09-08 01:03:24 -05:00 |
|
|
|
67617d7d69
|
also cull frozen_params in the params optimizer receives to reduce VRAM it consumes
|
2023-09-07 18:27:02 -05:00 |
|
|
|
8837bc34d7
|
added option to specify parameters to freeze per-model in YAML (because I need to see about committing atrocities with convering an AR into an AR+NAR)
|
2023-09-07 18:19:51 -05:00 |
|
|
|
c47fc3274e
|
added backwards compat flag
|
2023-09-07 17:12:17 -05:00 |
|
|
|
ab5134f385
|
tweaks and fixes
|
2023-09-07 17:08:38 -05:00 |
|
|
|
b2c2dec291
|
added homebrewed per-RVQ-bin embedding solutions
|
2023-09-07 16:48:02 -05:00 |
|
|
|
e7a67410d1
|
oops
|
2023-09-07 09:14:03 -05:00 |
|
|
|
712808494f
|
added support for optional prodigy optimizer (https://github.com/konstmish/prodigy) although it consumes a lot more VRAM per parameter
|
2023-09-06 20:33:16 -05:00 |
|
|
|
7ce06432fd
|
fixed the AR+NAR dual model, the resp_emb has to be split up (classifier might too)
|
2023-09-06 19:33:39 -05:00 |
|
|
|
100ca6b7d0
|
added option to use SGD optimizer through the YAML, added option to pass in additional optimizer parameters through the YAML, added experimental unified AR+NAR model (does not seem fruitful in testing)
|
2023-09-06 18:58:35 -05:00 |
|
|
|
451726fdd5
|
added ability to disable activation checkpointing through the YAML (it is very VRAM intensive at double layer size)
|
2023-09-05 15:38:21 -05:00 |
|
|
|
143aee7526
|
removed dedicated interleaved AR code
|
2023-09-03 22:47:03 -05:00 |
|
|
|
2f9cd0842f
|
merged dedicated interleaved AR code with the normal AR code
|
2023-09-03 22:46:08 -05:00 |
|
|
|
3a6bd50322
|
haha
|
2023-09-03 21:36:58 -05:00 |
|
|
|
c56ce033d9
|
work on an interleaved AR (spoiler: it does not work)
|
2023-09-03 21:27:58 -05:00 |
|
|
|
8a6c203277
|
added per-speaker samplers
|
2023-09-03 21:27:13 -05:00 |
|
|
|
81b05dabb9
|
accurate epoch metric is now reported (based on samples processed / length of dataset's paths, rather than naive assumptions)
|
2023-09-03 08:03:36 -05:00 |
|
|
|
922404285c
|
fixed segfault from tts-c task token exceeding being too big (inserted it in the hypothetical svc task token because in reality that is never ever going to be a feasible task to train against)
|
2023-09-02 19:25:43 -05:00 |
|
|
|
4613781e23
|
integrated plot script, added tts-c task token to help the model be able to mix between normal VALL-E and VALL-E continuous
|
2023-09-02 16:29:53 -05:00 |
|
|
|
71e68a8528
|
tweaked tts-continuous task
|
2023-09-02 13:39:17 -05:00 |
|
|
|
57db3ccfa8
|
shuffled VALL-E continuous as a task tts-c instead, logic fixes for it
|
2023-09-02 12:23:40 -05:00 |
|
|
|
2f06166ddd
|
cleanups
|
2023-09-01 21:33:51 -05:00 |
|
|
|
e40c0d34a0
|
somewhat got recurrent forward working (it's as accurate as chunkwise forward: it's not accurate at all), added option to use AMP instead of blanket setting the weight's dtype
|
2023-09-01 20:58:29 -05:00 |
|
|
|
2bc2d08b09
|
(need to verify) added modifying model size and config bool to align with VALL-E continuous' methodology
|
2023-09-01 17:19:34 -05:00 |
|
|
|
5c8694db8e
|
nasty bandaid if there's no validation dataset specified during training (for example, during finetunes)
|
2023-08-30 18:23:05 -05:00 |
|
|
|
7f4388e591
|
added total samples processed and tokens processed (len of text tokens + len of target response tokens)
|
2023-08-28 11:02:45 -05:00 |
|
|
|
87c4bfedba
|
added ability to mark models as disabled for training, and hotloading them for eval/validation (useful if training only one model, or training a model per GPU)
|
2023-08-27 12:26:12 -05:00 |
|
|
|
165a1154e0
|
Undo naive=False test flag, this shouldn't have made its way in
|
2023-08-26 22:00:43 -05:00 |
|
|
|
78378ed1ce
|
overhauled dataloading code to be marginally faster, mostly cleaned up, and can leverage a metadata json to help things out
|
2023-08-26 19:53:23 -05:00 |
|
|
|
16e0020901
|
disabled chunkwise_recurrent for 2x speed gains (I suppose it has been working the entire time, but I have not been properly grabbing things, and this might explain why the output is bad)
|
2023-08-25 19:50:19 -05:00 |
|
|
|
6455a2f9d7
|
I think I fixed a bug?
|
2023-08-24 23:33:36 -05:00 |
|
|
|
0517d620b8
|
fixes with the local backend
|
2023-08-24 17:05:56 -05:00 |
|
|
|
00ad4af651
|
updated draconian requirement for espeak-ng to be installed and the env var set to the dll for Windows
|
2023-08-24 14:57:01 -05:00 |
|
|
|
22904a8639
|
more oversights fixed because I've been using a cached dataloader forever now and didn't catch these problems
|
2023-08-24 10:25:33 -05:00 |
|
|
|
5873c27f1a
|
ops
|
2023-08-24 09:20:47 -05:00 |
|
|
|
501a857d5d
|
ops
|
2023-08-23 17:03:25 -05:00 |
|
|
|
4585824cd3
|
tweaks, including exporting on save/quit
|
2023-08-23 16:43:03 -05:00 |
|
|
|
d106598403
|
do not utilize diskcache if a config yaml is not loaded
|
2023-08-23 11:02:15 -05:00 |
|
|
|
524d289c9c
|
Forgot to re-add in setting the weight's dtype on model load
|
2023-08-22 22:57:23 -05:00 |
|
|
|
9c5a33bfd2
|
added repo with my weights so far
|
2023-08-22 13:09:44 -05:00 |
|
|
|
7b1b82e0e5
|
inferencing cleanup
|
2023-08-20 21:36:02 -05:00 |
|
|
|
a47029065b
|
I don't know if the lack of start/stop tokens being added was causing my inference tests to fail, but it seems better now
|
2023-08-20 19:21:54 -05:00 |
|
|
|
736c077282
|
ops
|
2023-08-20 13:42:18 -05:00 |
|
|
|
b105f6211e
|
added ability to export weights mid-training to avoid CBT to yank the weights while the training script is running
|
2023-08-20 13:39:58 -05:00 |
|
|
|
fc576010ce
|
wrapped saving the checkpoint in a try/catch so I can stop waking up to the damn trainer crashing because it ran out of disk space; I'd much rather it keep training to give me time to eventually clear up disk space rather than it silently restarting on its own
|
2023-08-20 06:29:17 -05:00 |
|
|
|
2d1a9f10c0
|
nightmare of spaghetti that might break compat; mechanism to increase RVQ bins of an existing model without retraining, keeps sampled proms/resps at max RVQ level and trim off excess levels according to what model receives them, some other things I already forgot (I really hope no one else has weights being baked right now)
|
2023-08-19 15:06:33 -05:00 |
|
|
|
f7f6d3bf6d
|
validated that SpeechX tasks cse and nse works, added a method to test each task by invoking python3 -m vall_e.data --action=tasks --tasks='sr,se,cse,nse'
|
2023-08-19 09:50:07 -05:00 |
|
|
|
6ca347e1e1
|
literally had a urethra moment before going to bed with a way to implement cse/nse tasks
|
2023-08-19 01:16:46 -05:00 |
|
|
|
8f42c578c9
|
setting up for allowing training for a partial amount of the speechx tasks (do NOT try this at home yet without a proper model, as performance is predecated on having a solid base vall-e model for the tasks
|
2023-08-19 00:16:08 -05:00 |
|
|
|
ae9d38aa31
|
forgot to have it pull from specified noise to the hdf5 dataset
|
2023-08-18 23:57:07 -05:00 |
|
|
|
77292c42f9
|
tested the training preparation for tasks ns, sr, and tse (I don't expect it to go well with only 2 RVQ bins)
|
2023-08-18 23:55:40 -05:00 |
|
|
|
bbb0563b3d
|
pseudocode polyfill stub some other flavor of working on adding the tasks
|
2023-08-18 22:22:13 -05:00 |
|
|
|
0b46c1e312
|
god I am inexperienced with retaining compat from previous weights, I hope no one actually has weights
|
2023-08-18 21:29:20 -05:00 |
|
|
|
508677fcd5
|
repaired auraloss loss calc during eval/val
|
2023-08-18 21:19:47 -05:00 |
|
|
|
fb4e816823
|
oops
|
2023-08-18 21:11:19 -05:00 |
|
|
|
2a71486cb6
|
preparing for SpeechX extensions
|
2023-08-18 20:58:07 -05:00 |
|
|
|
ced31fd9b7
|
removed the sampler as it's very misleading
|
2023-08-18 14:47:48 -05:00 |
|
|
|
8e7f900210
|
forgot the =
|
2023-08-17 19:07:59 -05:00 |
|
|
|
3ff7cf8341
|
maybe fix evaluation dataset not being capped to cfg.evaluation.size
|
2023-08-17 18:56:37 -05:00 |
|
|
|
ee58db746f
|
actually make the evaluation dataset shuffled for sample_type=speaker
|
2023-08-17 15:04:45 -05:00 |
|
|
|
18403a3523
|
maybe fixes eval dataloader not shuffling under distributed
|
2023-08-17 13:41:53 -05:00 |
|
|
|
03872b823f
|
why did I type rglob, another 10 bucks down the drain...
|
2023-08-17 00:11:29 -05:00 |
|
|
|
b5f247aa11
|
just nuked about 9 hours of progress because I didn't make sure it pruned only on the global leader
|
2023-08-16 23:37:52 -05:00 |
|
|
|
d7152fc7b9
|
added pruning of old checkpoints if specified (cfg.trainer.keep_last_checkpoints)
|
2023-08-16 20:12:12 -05:00 |
|
|
|
44c08d828e
|
added sample_type that samples from speakers to truly balance an epoch by speakers rather than the entire dataset and a sampler that tries to balance by speakers
|
2023-08-16 19:39:21 -05:00 |
|
|
|
599e47a813
|
might fix user inputted saving/quitting breaking when distributed
|
2023-08-15 23:52:20 -05:00 |
|
|
|
1e3e1d9315
|
tweaks
|
2023-08-15 21:58:16 -05:00 |
|
|
|
277c759ab1
|
fixed issue with non-distributed training, oops
|
2023-08-14 21:42:35 -05:00 |
|
|
|
5fa86182b5
|
oops
|
2023-08-14 10:50:40 -05:00 |
|
|
|
13571380be
|
made exporter make more sense
|
2023-08-13 22:56:28 -05:00 |
|
|
|
d7deaf6def
|
distributed training works now (hopefully)
|
2023-08-13 22:07:45 -05:00 |
|
|
|
2af09d0bef
|
fixed that mysterious discepancy between the reported losses (I am so freaking mad, my piss is boiling, I had to interrupt halfway through an epoch)
|
2023-08-05 15:25:41 -05:00 |
|
|
|
d1b9770d41
|
set model to eval when inferencing (very important)
|
2023-08-05 04:29:05 +00:00 |
|
|
|
d89568a96e
|
some fixes for the local framework
|
2023-08-05 03:22:15 +00:00 |
|
|
|
5970f254e3
|
some fixes for the local framework
|
2023-08-05 02:17:30 +00:00 |
|
|
|
012f54b7f1
|
another classic commit so i can copy it to another machine to gut out things and use the trainer bits for a side project that I should really get around to working on sooner than later
|
2023-08-04 14:21:30 -05:00 |
|
|
|
0a524f1d59
|
reticulating splines
|
2023-08-03 21:39:00 -05:00 |
|
|
|
608c1970eb
|
ops
|
2023-08-03 20:36:19 -05:00 |
|
|
|
c85101403f
|
big cleanup
|
2023-08-03 20:26:36 -05:00 |
|
|
|
2e03e5ac93
|
Fixed an issue with having fairseq installed at all will brick logging
|
2023-08-02 22:57:10 -05:00 |
|
|
|
f6597e2dfe
|
adjustments
|
2023-08-02 18:36:26 -05:00 |
|
|
|
0f9b81de75
|
oops
|
2023-08-02 18:12:36 -05:00 |
|
|
|
7a06b27a9c
|
Tweaks
|
2023-08-02 22:06:39 +00:00 |
|
|
|
bf8cedc9dd
|
Rewrite init
|
2023-08-02 21:53:35 +00:00 |
|
