|
|
0e995dbf2c
|
is this my last cope (falling back to explicit duration prediction, as this regression just won't go away) (also the smaller model was lobotomized because of my ROCm setup having a botched SDPA for who knows why)
|
2025-04-02 17:01:24 -05:00 |
|
|
|
7a0956863d
|
oops
|
2025-03-31 21:11:43 -05:00 |
|
|
|
a1184586ef
|
should never have trusted mse_loss, it never works
|
2025-03-31 20:59:13 -05:00 |
|
|
|
478aea0e8c
|
tweaks
|
2025-03-28 19:49:54 -05:00 |
|
|
|
6ae282e090
|
re-added noise dataloader sampler whatever for the old implementation's other tasks that require it
|
2025-03-28 15:07:06 -05:00 |
|
|
|
90b3509404
|
I'll just cope and say I cannot apply segmented attention masks to the smaller model as it's too trained on not doing it, and the regression came from dumb python aliasing rules
|
2025-03-27 13:27:51 -05:00 |
|
|
|
2fd82a7a22
|
cannot get segmented mask to actually work without gradients exploding (need to find a different way to do duration prediction...)
|
2025-03-27 00:51:41 -05:00 |
|
|
|
09e9438941
|
ugh
|
2025-03-25 23:24:01 -05:00 |
|
|
|
8641c87611
|
nothing could go wrong part 2 (reverted and rewrote commits since there was a nasty regression)
|
2025-03-25 23:06:16 -05:00 |
|
|
|
02a8bcbe29
|
fixed errant index error (although it makes me wonder if my segmented masking is still flawed)
|
2025-03-21 23:41:34 -05:00 |
|
|
|
d1d91295b3
|
add segmented sliding attention, also found a bug with prom-less segments in the attention mask generation.........
|
2025-03-21 19:05:49 -05:00 |
|
|
|
589cfb0e18
|
yuge speedup because of a dumb oversight
|
2025-03-20 17:39:41 -05:00 |
|
|
|
8068f24e35
|
cleaned up parallel nar, i think it's slightly faster but even the smallest model is still slower than ar+nar-len-llama-8...
|
2025-03-20 15:56:15 -05:00 |
|
|
|
9a7458cf17
|
fixed inferencing since I did delete the len_emb, some more notes on the model since it seems I just had bad experimental settings
|
2025-03-19 22:41:48 -05:00 |
|
|
|
61de653ad9
|
now causal training should work again
|
2025-03-19 14:20:19 -05:00 |
|
|
|
85b9dd47c1
|
ugh
|
2025-03-19 13:31:50 -05:00 |
|
|
|
81acd565b3
|
re-enable these
|
2025-03-18 20:59:33 -05:00 |
|
|
|
5479d2eacc
|
more tweaks to the new implementation (properly trim the len stuff to save some params, decoder to d_ffn expansion to 2 to maybe also make it faster, etc.)
|
2025-03-18 19:34:37 -05:00 |
|
|
|
9a8a8e3195
|
off by one bateman
|
2025-03-18 08:40:43 -05:00 |
|
|
|
0280e72257
|
ugh
|
2025-03-17 21:49:45 -05:00 |
|
|
|
b0dba9db07
|
this may bite me in the ass
|
2025-03-17 21:46:50 -05:00 |
|
|
|
2dfef693c4
|
comments for clarity
|
2025-03-16 11:30:23 -05:00 |
|
|
|
9cfbf94b1c
|
config-ify the len_loss_factor
|
2025-03-14 20:30:48 -05:00 |
|
|
|
ca8cc15271
|
more tweaks (vall_e.webui --yaml still breaks things, --model needs to deduce what audio backend now that im supporting other ones again // added easy top-sampler settings back for new implementation)
|
2025-03-14 20:18:25 -05:00 |
|
|
|
6ee505cffd
|
fixed dac
|
2025-03-12 23:17:27 -05:00 |
|
|
|
ba5f3d19b4
|
use the FSQ-targeted encoder/decodede whole-ly as it works for EnCodec too, as the RVQ-targeted encoder/decoder doesnt (and some notes)
|
2025-03-12 22:47:19 -05:00 |
|
|
|
2ccf1b5740
|
actually do duration prediction
|
2025-03-11 22:14:54 -05:00 |
|
|
|
5c512717a6
|
len prediction for new model (and remove logit normalization since it kills inferencing)
|
2025-03-11 20:33:09 -05:00 |
|
|
|
5f98543d4d
|
ughh
|
2025-03-10 21:18:57 -05:00 |
|
|
|
8ac03aac8a
|
ugh
|
2025-03-10 21:14:56 -05:00 |
|
|
|
5670fcb23f
|
hopefully the final tweaks needed for this bastard of a model
|
2025-03-10 20:59:11 -05:00 |
|
|
|
00d1fed217
|
another optimization (within the dataloader because the similar utterance sampler was mondo slow)
|
2025-03-08 17:10:50 -06:00 |
|
|
|
5e9d1a5302
|
one more time one more time (this normalization isn't a spook)
|
2025-03-07 19:32:42 -06:00 |
|
|
|
93044829af
|
one more time (could have sworn i tested it with batch size > 1)
|
2025-03-07 19:14:33 -06:00 |
|
|
|
6cea840710
|
oops
|
2025-03-07 18:57:25 -06:00 |
|
|
|
dbd34b6430
|
add specialized calc_loss because schizo
|
2025-03-07 18:44:11 -06:00 |
|
|
|
8d848ed549
|
handle case of dropping cond for segment mask
|
2025-03-07 14:11:58 -06:00 |
|
|
|
89e52b9877
|
ugh
|
2025-03-07 13:55:57 -06:00 |
|
|
|
6afc2b7526
|
gut feeling to change the attention mask
|
2025-03-07 13:51:59 -06:00 |
|
|
|
ec87308d75
|
final tweaks before training this meme 44khz model for the 3rd time
|
2025-03-06 15:31:15 -06:00 |
|
|
|
5cd71ef238
|
QoL so I can stop having to manually inject different configs
|
2025-03-06 14:48:14 -06:00 |
|
|
|
0d809561c6
|
accuracy k=1 and k=80 because im probably dumb for k=10 as the default since it does not represent any usecase
|
2025-03-05 16:35:34 -06:00 |
|
|
|
2fb2b732fc
|
wow that was fast
|
2025-03-04 23:17:18 -06:00 |
|
|
|
462f71e2f7
|
ugh
|
2025-03-04 14:57:00 -06:00 |
|
|
|
1cd24f3381
|
a birdie tells me i should probably use a different optimizer (also preliminary support for native sparse attention but I don't know if I'll use it)
|
2025-03-04 14:53:02 -06:00 |
|
|
|
0451f75e33
|
now that the new model seems a little more promising, i can re-document things non-cynically
|
2025-03-03 13:21:41 -06:00 |
|
|
|
3f1070f575
|
tweaks
|
2025-03-02 22:36:25 -06:00 |
|
|
|
17094b8002
|
reticulating splines
|
2025-03-01 17:48:51 -06:00 |
|
|
|
ddc49c89c5
|
the learning rate scheduler pill is a tough pill to swallow
|
2025-02-28 22:12:19 -06:00 |
|
|
|
b97faa8173
|
fixes...
|
2025-02-28 18:53:07 -06:00 |
|
