a5c21d65d2added automatically loading default YAML if --yaml is not profided (although I think it already does this by using defaults), default YAML will use local backend + deepspeed inferencing for speedupsmrq2024-06-19 18:49:39 -0500
f4fcc35aa8fixed it breaking on subsequent utterances through the web UI from latents being on the CPUmrq2024-06-19 18:26:15 -0500
96b74f38efsampler and cond_free selectable in webUI, re-enabled cond_free as default (somehow it's working again)mrq2024-06-19 17:12:28 -0500
73f271fb8aadded automagic offloading models to GPU then CPU when theyre done during inferencemrq2024-06-19 17:01:05 -0500
5d24631bfbdon't pad output mel tokens to speed up diffusion (despite copying it exactly from tortoise)mrq2024-06-19 15:27:11 -0500
849de13f27added tqdm bar for ARmrq2024-06-19 15:00:14 -0500
99be487482backported old fork features (kv_cache (which looking back seems like a spook), ddim sampling, etc)mrq2024-06-19 14:49:24 -0500
268ba17485crammed in HF attention selection mechanisms for the ARmrq2024-06-19 10:21:43 -0500
e5136613f5semblance of documentation, automagic model downloading, a little saner inference results foldermrq2024-06-19 10:08:14 -0500
6c2e00ce2aload exported LoRA weights if exists (to-do: make a better LoRA loading mechanism)mrq2024-06-18 21:46:42 -0500
cbd3c95c42possible speedup with one simple trick (it worked for valle inferencing), also backported the voice list loading from aivcmrq2023-08-20 22:32:01 -0500
9afa71542blittle sloppy hack to try and not load the same model when it was already loadedmrq2023-08-11 04:02:36 +0000
c90ee7c529removed kludgy wrappers for passing progress when I was a pythonlet and didn't know gradio can hook into tqdm outputs anywaysmrq2023-05-04 23:39:39 +0000
086aad5b49quick hotfix to remove offending codesmell (will actually clean it when I finish eating)mrq2023-05-04 22:59:57 +0000
8618922a33Implement correct XPU device count Forgot to do thata-One-Fan2023-05-04 21:14:07 +0300
04b7049811freeze numpy to 1.23.5 because latest version will moan about deprecating complexmrq2023-05-04 01:54:41 +0000
2f7d9ab932disable BNB for inferencing by default because I'm pretty sure it makes zero differences (can be force enabled with env vars if you'r erelying on this for some reason)mrq2023-04-29 00:38:18 +0000
f025470d60Merge pull request 'Update tortoise/utils/devices.py vram issue' (#44) from aJoe/tortoise-tts:main into main
mrq
2023-04-12 19:58:02 +0000
d1ad634ea9added japanese preprocessor for tokenizermrq2023-03-17 20:03:02 +0000
af78e3978adeduce if preprocessing text by checking the JSON itself insteadmrq2023-03-16 14:41:04 +0000
e201746eebadded diffusion_model and tokenizer_json as arguments for settings editingmrq2023-03-16 14:19:24 +0000
1f674a468fadded flag to disable preprocessing (because some IPAs will turn into ASCII, implicitly enable for using the specific ipa.json tokenizer vocab)mrq2023-03-16 04:33:03 +0000
42cb1f3674added args for tokenizer and diffusion model (so I don't have to add it later)mrq2023-03-15 00:30:28 +0000
65a43deb9ewhy didn't I also have it use chunks for computing the AR conditional latents (instead of just the diffusion aspect)mrq2023-03-14 01:13:49 +0000
97cd58e7ebmaybe solved that odd VRAM spike when doing the clvp passmrq2023-03-12 12:48:29 -0500
3dd5cad324reverting additional auto-suggested batch sizes, per mrq/ai-voice-cloning#87 proving it in fact, is not a good ideamrq2023-03-07 19:38:02 +0000
cc36c0997cdidn't get a chance to commit this this morningmrq2023-03-07 15:43:09 +0000
7f98727ad5added option to specify autoregressive model at tts generation time (for a spicy feature later)mrq2023-03-06 20:31:19 +0000
6fcd8c604fmoved bigvgan model to a huggingspace repomrq2023-03-05 19:47:22 +0000
0f3261e071you should have migrated by now, if anything breaks it's on (You)mrq2023-03-05 14:03:18 +0000
06bdf72b89load the model on CPU because torch doesn't like loading models directly to GPU (it just follows the default vocoder loading behavior)mrq2023-03-03 13:53:21 +0000
a9de016230added storing the loaded model's hash to the TTS object instead of relying on jerryrig injecting it (although I still have to for the weirdos who refuse to update the right way), added a parameter when loading voices to load a latent tagged with a model's hash so latents are per-model nowmrq2023-03-02 00:44:42 +0000
7b839a4263applied the bitsandbytes wrapper to tortoise inference (not sure if it matters)mrq2023-02-28 01:42:10 +0000
7cc0250a1aadded more kill checks, since it only actually did it for the first iteration of a loopmrq2023-02-24 23:10:04 +0000
de46cf7831adding magically deleted files back (might have a hunch on what happened)mrq2023-02-24 19:30:04 +0000
2c7c02eb5cmoved the old readme back, to align with how DLAS is setup, sortamrq2023-02-19 17:37:36 +0000
eca61af016actually for real fixed incrementing filenames because i had a regex that actually only worked if candidates or lines>1, cuda now takes priority over dml if you're a nut with both of them installed because you can just specify an override anywaysmrq2023-02-16 01:06:32 +0000
ec80ca632badded setting "device-override", less naively decide the number to use for results, some other thingmrq2023-02-15 21:51:22 +0000
mrq
Jarod Mica
HarkonCollider
mrq
ken11o2
a-One-Fan
aJoe
NtTestAlert
deviandice