|
|
5fe01ffc6c
|
more notes / re-enabled top-k/p samplers for new implementation
|
2025-04-19 14:04:34 -05:00 |
|
|
|
d9e18037cc
|
new implementation tweaks and fixes to make it actually better (there were a lot of badwrong things being done that harmed the output quality, will evaluate the model further)
|
2025-04-18 20:36:44 -05:00 |
|
|
|
98d1d8cb1e
|
added some more notes, tweaks (RIP DAC, it's over)
|
2025-04-17 20:24:40 -05:00 |
|
|
|
6d42c9ae23
|
how foolish of me, not having a softmax as float32 (maybe addresses an emergent regression where bfloat16 training shits the bed where float16+loss scaling doesnt)
|
2025-04-07 22:51:52 -05:00 |
|
|
|
d6cd848c32
|
goodbye nvidia/audio-codec-44khz, crossed fingers for DAC again
|
2025-04-06 21:05:29 -05:00 |
|
|
|
2e93438867
|
reintroduced sampler_type = speaker because I think this might salvage the nemo model to have better speaker similarities
|
2025-04-03 19:01:10 -05:00 |
|
|
|
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 |
|
|
|
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 |
|
|
|
4d777b5618
|
add remark that segmented attention actually might be broken (for some reason this only emerged recently, need to investigate)
|
2025-03-26 12:08:47 -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 |
|
|
|
aa8b32d97e
|
added more notes (although I could have sworn I have had more notes that i can't recall)
|
2025-03-25 18:53:06 -05:00 |
|
|
|
df5b870908
|
added remark about not using sliding attention
|
2025-03-22 12:44:34 -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 |
|
|
|
81acd565b3
|
re-enable these
|
2025-03-18 20:59:33 -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 |
|
|
|
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 |
|
|
|
5c512717a6
|
len prediction for new model (and remove logit normalization since it kills inferencing)
|
2025-03-11 20:33:09 -05:00 |
|
|
|
5cd71ef238
|
QoL so I can stop having to manually inject different configs
|
2025-03-06 14:48:14 -06:00 |
|
|
|
2fb2b732fc
|
wow that was fast
|
2025-03-04 23:17:18 -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 |
|
|
|
4afa4ccce5
|
at wits end (parhaps the semantic token approach is the toughest pill to swallow)
|
2025-03-01 21:03:25 -06:00 |
|
|
|
a174c33db6
|
a gorillionth time's the charm (aka: the encoder/decoder pill is a tough pill to swallow)
|
2025-02-28 17:56:50 -06:00 |
|
|
|
eff180248c
|
decoupled llama backend to avoid any funny changes from transformers, removed other backends since i dont think i'll ever bother using them
|
2025-02-27 19:00:37 -06:00 |
|
|
|
95da4e9405
|
made muon actually work by actually utilizing param groups (thanks APOLLO for reminding me this is the sane way to handle this split)
|
2025-02-26 10:39:13 -06:00 |
|
|
|
92139b6da9
|
additional cruft, added a note in documentation to be aware of NUMA node topology when running vall_e.emb.process with more than one process
|
2025-02-18 19:56:30 -06:00 |
|
|
|
0dc49ef4d5
|
documentation update while I wait for more audio (between 4 and 8 seconds per utterance) quantize for nvidia/audio-codec-44khz (I was foolish to think I can get something servicable with just 4 seconds max for an utterance)
|
2025-02-15 17:42:06 -06:00 |
|
|
|
04fef5dad5
|
agony
|
2025-02-12 00:18:24 -06:00 |
|
|
|
1c0ed6abac
|
added notes on this unfruitful experiment
|
2025-02-11 16:21:43 -06:00 |
|
|
|
9fa87c417a
|
added option to use raw text rather than the IPA phonemes (it requires a model trained on raw text)
|
2025-01-06 00:10:43 -06:00 |
|
|
|
9b0d2ccbe1
|
|
2024-12-26 21:42:17 -06:00 |
|
|
|
59bf6b8b33
|
exposed additional task (ns, sr, vc) (vc is experimental)
|
2024-12-20 11:15:29 -06:00 |
|
|
|
8515038968
|
imagine my disappointment when the epoch finished just for it to throw an exception
|
2024-12-16 18:28:01 -06:00 |
|
|
|
f41251f648
|
more fixes for local engine backend
|
2024-12-12 14:38:42 -06:00 |
|
|
|
8568a93dad
|
added WER/SIM-O metrics, added APOLLO but I need to test it
|
2024-12-10 20:13:21 -06:00 |
|
|
|
a6c745bafb
|
chinese (mandarin?) support added (I guess I don't need pinyin, but tone markers are handled), korean validated, vocab adjusted
|
2024-12-09 14:26:19 -06:00 |
|
|
|
a032ff588f
|
doc update, added automatically deducing language from a given text, also checks if the input is already phonemized text to allow direct control without being cringe (procrastinating adding WER/SIM-O)
|
2024-12-07 22:34:25 -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 |
|
|
|
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 |
|
|
|
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 |
|
|
|
efeb55e1b7
|
documentation update
|
2024-11-19 19:19:34 -06:00 |
|
|
|
190a917b3e
|
I did it.
|
2024-11-19 12:24:33 -06:00 |
|
|
|
5ba80686e1
|
two weeks of agony concludes
|
2024-11-18 21:29:28 -06:00 |
|
