|
|
f97e8b0c7f
|
ACTUALLY do KD-loss because of an oversight with masked_select outputting 1D tensors that get softmax'd in total
|
2024-12-07 09:52:51 -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 |
|
|
|
4e21df8092
|
oops
|
2024-12-04 21:24:22 -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 |
|
|
|
31ab90d84a
|
cringe code to convert to LlamaForCausalLM-happy weights + tokenizer dict (still need to write logic to actually use these weights for proper inferencing)
|
2024-12-03 10:18:58 -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 |
|
|
|
3fc0540f49
|
m
|
2024-11-21 15:07:46 -06:00 |
|
|
|
6845c447c9
|
added more harvard sentences to load from a text file
|
2024-11-21 13:18:11 -06:00 |
|
|
|
2a084544e8
|
moved duration padding for NAR-len to be a scalar instead (since it seems longer utterances need it much more so than shorter utterances)
|
2024-11-21 13:04:07 -06:00 |
|
|
|
6aee08f9c0
|
moved stuff in the web UI around (un-experimented the max NAR-len steps because its kind of important to adjust this value for better sounding audio / quicker generated audio)
|
2024-11-20 20:37:33 -06:00 |
|
|
|
dfdba3f190
|
oops
|
2024-11-20 19:21:03 -06:00 |
|
|
|
cd6e9ba2f2
|
oops
|
2024-11-20 16:27:51 -06:00 |
|
|
|
1a73ac6a20
|
I cannot believe it's not actually called Wand DB (added wandb logging support since I think it would have been a much better way to look at my metrics)
|
2024-11-20 16:10:47 -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 |
|
|
|
db64e6cb59
|
dependency updates (gradio 5.x now works on my machine)
|
2024-11-20 12:33:01 -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 |
|
|
|
4a71981456
|
normalize sampler index by batch size (if not using batched sampler), add option to cap out utterances for a speaker, some other things
|
2024-11-18 12:46:50 -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 |
|
|
|
ef05c951ff
|
adjust fp16 loss scaling since I fried a model overnight when it hit 8K scale
|
2024-11-14 09:23:52 -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 |
|
|
|
29e45be0b4
|
tweaks to bucket sampling
|
2024-11-13 11:09:24 -06:00 |
|
|
|
b2eca271a8
|
ugh
|
2024-11-13 10:35:44 -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 |
|
|
|
976ee87f6f
|
resume iteration step in tqdm trainer, warn to logger if the sampler state dict was invalidated
|
2024-11-13 09:09:28 -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 |
|
|
|
caf721c67b
|
set it to zero because it'll make the stop token hide more often than not
|
2024-11-12 22:30:50 -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 |
|
|
|
ec92613847
|
actually pass input prompt length size to inference
|
2024-11-11 20:39:48 -06:00 |
|
|
|
b1df6a7bed
|
reverted rep pen sampler due to a regression
|
2024-11-11 20:35:08 -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 |
|
|
|
354f8e059d
|
store dataset hash alongside state dict so it can be ignored if mismatched
|
2024-11-11 18:16:56 -06:00 |
|
|
|
f7b8b1e825
|
dropped subtrain dataloader since its useless to duplicate
|
2024-11-11 17:00:49 -06:00 |
|
|
|
cf9df71f2c
|
use homwbrewed caching system for dataloader paths / durations (I'm pretty sure I am now triggering OOM killers with my entire dataset used)
|
2024-11-11 16:32:08 -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 |
|
|
|
a3bc26f7ec
|
ugh
|
2024-11-06 23:16:28 -06:00 |
|
|
|
d606a693ff
|
eval fix for nar-len
|
2024-11-06 23:14:16 -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 |
|
|
|
bcabde3454
|
more notes
|
2024-11-06 13:51:28 -06:00 |
|
|
|
bfc5e1d723
|
agony
|
2024-11-05 22:30:49 -06:00 |
|
|
|
aefe8fcdad
|
UGH
|
2024-11-05 22:13:58 -06:00 |
|
|
|
556d9db0d5
|
web UI support for HF ZeroGPU
|
2024-11-05 21:38:02 -06:00 |
|
|
|
e58a9469a3
|
move layerskip to experimental settings.......
|
2024-11-05 20:37:06 -06:00 |
|
|
|
bbc2de3713
|
ugh
|
2024-11-05 11:50:05 -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 |
|
