|
|
8515038968
|
imagine my disappointment when the epoch finished just for it to throw an exception
|
2024-12-16 18:28:01 -06:00 |
|
|
|
4a65ac9eb7
|
oops
|
2024-12-15 17:21:51 -06:00 |
|
|
|
9a62e3b824
|
APOLLO cringe (doesn't want to work with deepspeed)
|
2024-12-12 00:31:58 -06:00 |
|
|
|
cddf8ca814
|
sort batches to try and reduce number of padded tokens in batched inference (also commented out F5 samples getting added to the demo page because I would have to regenerate them)
|
2024-12-11 22:45:38 -06:00 |
|
|
|
6468e5d124
|
lol
|
2024-12-11 19:10:32 -06:00 |
|
|
|
3ef8894290
|
oops
|
2024-12-08 15:24:21 -06:00 |
|
|
|
1d460b9fe3
|
logic fixes, I feel like output is better? (also NAR can have a temperature, I imagine it couldn't because it was having a causal masked passed to it for the longest time before I caught it a month ago)
|
2024-12-08 14:52:47 -06:00 |
|
|
|
5d80a2d0d4
|
fixed NAR-len issues with non-english maybe (langs weren't being passed), added interface to inference in batches through tts.batched_inference (no support for rolling context/prefixes because there's no way to do that), demo page uses batched inferencing now
|
2024-12-07 19:21:05 -06:00 |
|
|
|
61ed662856
|
ACTUALLY actually fix KD-loss (the -inf in the logits was caused by cringecode)
|
2024-12-07 12:31:54 -06:00 |
|
|
|
34a66e1052
|
agnostified KD
|
2024-12-06 23:53:46 -06:00 |
|
|
|
953d3eb030
|
ugh
|
2024-12-06 22:35:30 -06:00 |
|
|
|
42fafbaaca
|
actually fixed knowledge distillation because of errant -inf logits causing problems and needed to be filtered (and splitting text language / output audio language because it helps)
|
2024-12-06 21:55:20 -06:00 |
|
|
|
23d402bf01
|
added knowledge distillation in the trainer (sadly it is not agnostic because of the grave mistake of further processing the batch within the forward pass, so subsequent calls do not match......)
|
2024-12-05 23:05:52 -06:00 |
|
|
|
93d27be539
|
rolling context finally (use last N utterances as the prefix for the next gen), option to split input text prompt by sentences instead of lines (or no splitting)
|
2024-12-04 20:31:44 -06:00 |
|
|
|
9dff68c0c5
|
NAR-len tweaks (remasks a small amount of tokens per step, it seems to help with reducing the number of steps needed some of the time?, disable CFG for the first half to speed things up)
|
2024-12-04 09:30:29 -06:00 |
|
|
|
cf97560e70
|
minimum CFG of 3 for NAR-len because it seems the model will auto-default to NAR-len now
|
2024-12-03 19:40:05 -06:00 |
|
|
|
ca31da0a95
|
sageattn (forgot to bother with testing this the other day, seems ifne)
|
2024-12-03 15:14:57 -06:00 |
|
|
|
84a05acb6d
|
touch ups in docs
|
2024-12-02 19:10:42 -06:00 |
|
|
|
dcaf38b359
|
fixed training tqdm being stubborn
|
2024-11-23 09:45:23 -06:00 |
|
|
|
41d7c30ea5
|
added much cleaner non-causal mask generation
|
2024-11-22 19:43:32 -06:00 |
|
|
|
c99a74e834
|
actually generate a causal mask because it seems sometimes it does not actually generate one because it makes assumptions
|
2024-11-22 18:30:24 -06:00 |
|
|
|
ccee5fc11c
|
that was actually all pointless since sdpa always had an attention mask fed to it and does not need is_causal to implicitly generate one
|
2024-11-22 16:51:50 -06:00 |
|
|
|
4aa685e749
|
what has science done
|
2024-11-22 16:45:40 -06:00 |
|
|
|
147219a5e0
|
huge oversight in the attention masking......... (i realized I have not been providing a non-causal mask to non-causal tasks)
|
2024-11-22 13:44:43 -06:00 |
|
|
|
24d888c47c
|
temporarily dropping support for xformers because it's breaking when using an attention mask (which i dont remember commenting it out when being passed), default to not use wandb because it's being a pain when doing tests and not actual sessionsS)
|
2024-11-22 11:29:12 -06:00 |
|
|
|
8aafae91fd
|
dont use timeembedding
|
2024-11-21 23:14:52 -06:00 |
|
|
|
2cef97e43f
|
cleanup
|
2024-11-21 23:08:43 -06:00 |
|
|
|
67f7bad168
|
added mixed modality AR+NAR-len to generate a short prefix through the AR, then inference with said prefix through the NAR-len (need to experiment with it more to ensure that the masked off tokens are the only tokens getting updated)
|
2024-11-20 14:22:12 -06:00 |
|
|
|
b1369e7824
|
better modality selection (pick AR+NAR by default for the ar+nar model, pick NAR-len by default for the nar-len model), lowered default CFG because it makes the AR+NAR output sped up (but can't be too low since it's required for the NAR-len)
|
2024-11-19 18:51:17 -06:00 |
|
|
|
190a917b3e
|
I did it.
|
2024-11-19 12:24:33 -06:00 |
|
|
|
0e621354e7
|
cleaned up classifier-free guidance logit processing (in order to try and cope with a bad nar-len model)
|
2024-11-19 10:30:05 -06:00 |
|
|
|
5ba80686e1
|
two weeks of agony concludes
|
2024-11-18 21:29:28 -06:00 |
|
|
|
2b29790173
|
oops
|
2024-11-18 14:12:26 -06:00 |
|
|
|
6cfdf94bf9
|
swap priority to use nar-len if available, added notes
|
2024-11-18 09:40:04 -06:00 |
|
|
|
069b27570f
|
set option to set training masking ratio (I don't think for tts a fixed masking ratio is beneficial since the magic of the AR+NAR is being able to still reference the prior sequence of tokens for predicting things)
|
2024-11-17 17:04:07 -06:00 |
|
|
|
88d840218d
|
default set cfg strength to 3.0 since the reference model is updated
|
2024-11-17 10:23:40 -06:00 |
|
|
|
a3e1fa3518
|
ugh
|
2024-11-17 09:28:33 -06:00 |
|
|
|
23fdba0c98
|
tweaks and changes
|
2024-11-16 15:49:06 -06:00 |
|
|
|
2fbeacfe92
|
ugh
|
2024-11-14 22:18:33 -06:00 |
|
|
|
39096f8ff3
|
redid loss calculation to be cleaner, and position ID generation, and other things (I might need to train the NAR-len from scratch and not resume from an existing checkpoint.........)
|
2024-11-14 22:17:47 -06:00 |
|
|
|
e412e98125
|
ugh
|
2024-11-14 07:34:22 -06:00 |
|
|
|
c00fc18b62
|
actually use the right embedding for nar-len
|
2024-11-13 18:04:04 -06:00 |
|
|
|
3ea8a610d6
|
fix STT
|
2024-11-13 14:27:15 -06:00 |
|
|
|
910033343c
|
overhauled how the right resp level / classifier gets picked to avoid cringemath
|
2024-11-13 13:31:17 -06:00 |
|
|
|
269648605e
|
move NAR-len rvq level 0 to separate embedding
|
2024-11-13 11:38:58 -06:00 |
|
|
|
be83ddabaa
|
better causal-ness for split loss calc, and also do masking for NAR-len for it
|
2024-11-13 10:17:52 -06:00 |
|
|
|
6b76419123
|
ugh
|
2024-11-13 09:54:20 -06:00 |
|
|
|
ad7cfffc00
|
NAR-len RVQ-0 was being trained causally.............
|
2024-11-13 09:43:50 -06:00 |
|
|
|
8286aa54c8
|
do not pass timestep token/embedding since it doesn't seem to matter at all after all, fixed training masking rate to 80% because a paper said so
|
2024-11-13 09:07:10 -06:00 |
|
|
|
0f2584eba7
|
new meme sampler PogChamp new meme sampler PogChamp (it sort of helps?)
|
2024-11-12 22:30:09 -06:00 |
|
|
|
663f07038d
|
haha... (do not create a token dropout/noise mask when not training (this sadly didnt fix NAR-len output))
|
2024-11-12 16:41:58 -06:00 |
|
|
|
b09328069e
|
actually do CFG sampling for base AR+NAR tasks
|
2024-11-12 13:42:39 -06:00 |
|
|
|
2495a7ef67
|
Fixed STT in the web UI
|
2024-11-12 12:49:53 -06:00 |
|
|
|
8927bad7bc
|
actually fixed rep pen (for ar and nar, it seems to help with nar unmasking)
|
2024-11-11 21:40:19 -06:00 |
|
|
|
b1f4db39c8
|
threw in CFG sampling for normal model as well to experiment with
|
2024-11-11 20:27:38 -06:00 |
|
|
|
2f56696506
|
overhauled inference/sampler kwargs to stop being a bloated mess
|
2024-11-11 20:21:16 -06:00 |
|
|
|
a748e223ce
|
tweaks
|
2024-11-11 12:40:41 -06:00 |
|
|
|
48490757da
|
fixes
|
2024-11-10 20:37:50 -06:00 |
|
|
|
9def34cd66
|
lol
|
2024-11-10 12:48:41 -06:00 |
|
|
|
9cb0b6901b
|
unified nar.py into ar_nar.py
|
2024-11-10 12:19:48 -06:00 |
|
|
|
a9d2faf2d7
|
all I can do now until I wait for the model to (re)train for pure NAR
|
2024-11-09 22:57:34 -06:00 |
|
|
|
ad7e290a5e
|
ugh (ROCm seems to silently clamp any token value >= logits.shape[-1] for loss calculation, while cuda will throw an assert, making it hard to find this dumb fuckup)
|
2024-11-09 19:40:02 -06:00 |
|
|
|
943fe70c10
|
I don't know why this fixes an assert thrown but it does
|
2024-11-09 19:04:13 -06:00 |
|
|
|
f50d92ba6c
|
Almost made a mistake
|
2024-11-09 18:12:54 -06:00 |
|
|
|
c6a38693a2
|
This better work
|
2024-11-09 18:04:59 -06:00 |
|
|
|
8b3d1cf70a
|
Something's Wrong
|
2024-11-09 15:07:43 -06:00 |
|
|
|
dcd5fecff3
|
some cleanup while I wait for the NAR-len to train to an acceptable state (currently it performs okay, but only on audo after 3 seconds or so)
|
2024-11-09 12:12:46 -06:00 |
|
|
|
69b0b3b854
|
set timestep tensor to whatever the time embedding's dtype is because it'll gripe under amp
|
2024-11-09 00:11:16 -06:00 |
|
|
|
5a09a5f6e9
|
I forgot about the time embedding...
|
2024-11-08 22:46:26 -06:00 |
|
|
|
811b15d280
|
I suppose I just have a shit training method since the sampler is as solid as I can get it...............
|
2024-11-08 22:05:41 -06:00 |
|
|
|
13b54953bd
|
agony
|
2024-11-08 13:34:39 -06:00 |
|
|
|
c127c4e488
|
'borrowed' a sampling scheduler for NAR-len's RVQ level 0 (better than before, but still not good enough)
|
2024-11-07 21:19:14 -06:00 |
|
|
|
e108c54daf
|
new NAR-len training paradigm......
|
2024-11-07 11:32:11 -06:00 |
|
|
|
ed174c589e
|
ugh
|
2024-11-07 09:19:21 -06:00 |
|
|
|
d13ab00ad8
|
one more note
|
2024-11-07 09:11:21 -06:00 |
|
|
|
5698188824
|
あたしって、ほんとバカ
|
2024-11-07 09:10:18 -06:00 |
|
|
|
77ff23e319
|
repeat extend the prom to fill the initial tokens for nar-len (it somewhat works, the model just needs to train more)
|
2024-11-06 23:29:53 -06:00 |
|
|
|
105ed51159
|
I guess I'll fall for the NAR-len meme again (I don't know where my previous weights are, so I need to train it again to test something)
|
2024-11-06 19:17:12 -06:00 |
|
|
|
aefe8fcdad
|
UGH
|
2024-11-05 22:13:58 -06:00 |
|
|
|
9e65e05e83
|
more windows specific fixes, limit gradio to <5.0.0 on linux (it works on windows, but not on my linux machine tm)
|
2024-11-04 18:00:33 -06:00 |
|
|
|
c83670c38c
|
Windows specific fixes (to-do: find libespeak-ng.dll automatically because it cannot be trusted to do it by default)
|
2024-11-03 19:19:15 -06:00 |
|
|
|
d229725c76
|
more adjustments (adjustments of early-exit entropy/varentropy thresholds, default rep pen being 1.5, experimental refine-on-stop, etc.)
|
2024-11-03 18:31:28 -06:00 |
|
|
|
aee08b7307
|
changed layerskip float16 training warning (since it didnt seem to fry on my 4xV100 system)
|
2024-11-03 09:58:29 -06:00 |
|
|
|
3826f9bae4
|
saner mask creation? (it doesnt matter, kv cache wont work)
|
2024-11-02 21:00:21 -05:00 |
|
|
|
ded746e157
|
very, very naive layerskip speculative sampling (it just checks if the current layer's state is good enough)
|
2024-11-02 11:49:05 -05:00 |
|
|
|
ec79230965
|
shuffled web UI options hidden by cfg.experimental to its own tab, expose early exit selection to inferencing (it kinda works naively, still need to implement self-speculation)
|
2024-11-01 21:30:06 -05:00 |
|
|
|
fb8faa295b
|
actually float16(+AMP) and layerskip is bad and will kill the model......
|
2024-11-01 18:36:44 -05:00 |
|
|
|
9b6c57bc57
|
third time's the charm (for some reason it escaped me that I should treat early exit loss as an aux_loss to be used with the normal loss, as if I was training a MoE's router)
|
2024-11-01 12:50:37 -05:00 |
|
|
|
76ebef45dc
|
off-by-one...
|
2024-10-31 13:24:48 -05:00 |
|
|
|
b63293cbbe
|
ugh
|
2024-10-30 22:49:11 -05:00 |
|
|
|
a22534e8f4
|
layer skip training implemented (need to gut the inferencing from the repo, and to actually see if the model can benefit from this)
|
2024-10-30 20:05:45 -05:00 |
|
|
|
ccf71dc1b6
|
added option to load from a model state dict directly instead of a yaml (to-do: do this for LoRAs too), automatically download the default model if none is provided
|
2024-10-25 22:15:15 -05:00 |
|
|
|
a96f5aee32
|
adjusted how i want to pass eval kwargs
|
2024-10-25 20:38:09 -05:00 |
|
|
|
92e6bff6dc
|
actually ar temp 0.5 with rep pen 1.125 seems to have the benefits of better outputs without it degrading some of the time but not all the time
|
2024-10-23 00:03:35 -05:00 |
|
|
|
8920e5e86b
|
actually have beam_width in the webUI work
|
2024-10-22 22:06:22 -05:00 |
|
|
|
910571ad34
|
too brainlet to diagnose why low temp / greedy sampling is randomly unstable some of the time
|
2024-10-22 20:13:54 -05:00 |
|
|
|
8eb9a4056b
|
modified default arguments (ar temp = 0 and rep pen = 1.125 seems to be stable, at least given the few things i tested), do not pass top k/top p/min p to NAR even though technically none of those things should matter when greedy sampling
|
2024-10-22 18:12:39 -05:00 |
|
|
|
1a02cd5bce
|
modify demo template to say F5 instead of YourTTS, swap LoRA comparison around to make the lora'd the base file, and the no-lora the suffix'd file
|
2024-10-21 19:52:02 -05:00 |
|
|
|
71731ed785
|
added prefixing with silence (was to test something, currently hidden under cfg.experimental=True)
|
2024-10-18 17:19:52 -05:00 |
|
|
|
fc8dfd8617
|
made greedy AR sampling viable (and preferable), with caveats (per comment in vall_e.models.ar_nar)
|
2024-10-18 16:55:00 -05:00 |
|
