Brandon Thomas
ea761a76b9
improved deployment w liveness + readiness probes
2020-01-06 19:33:12 -05:00
Brandon Thomas
374e1a477d
Server with Torch + Libtorch; Model load via env
2020-01-06 18:18:39 -05:00
Brandon Thomas
da29a1f403
This deployment works, finally
2020-01-06 04:48:07 -05:00
Brandon Thomas
a8858f2375
argument for address binding
2020-01-06 04:24:53 -05:00
Brandon Thomas
539570e079
k8s deployment, change network interface binding
2020-01-06 03:06:05 -05:00
Brandon Thomas
d5f4d229ff
move workflow (why didn't wysiwyg create here?)
2020-01-06 00:44:30 -05:00
Brandon Thomas
601fcb525d
YAML config for github action to publish to GPR
2020-01-06 00:43:19 -05:00
Brandon Thomas
de4cf5c174
Docker container. Breaks tacotron_melgan (haven't ported to actix)
2020-01-06 00:18:33 -05:00
Brandon Thomas
9f73582091
Adding skeleton of deployable service
2020-01-05 05:47:09 -05:00
Brandon Thomas
6fdbaa2ad1
this works
2020-01-05 03:29:17 -05:00
Brandon Thomas
91dea21a2f
This works!!
2020-01-04 05:15:30 -05:00
Brandon Thomas
914631c72c
It works! Wrong output sample rate - 20000
2020-01-04 05:11:51 -05:00
Brandon Thomas
edd521f2fd
exact audio tensor match with python tacotron-melgan (wav audio still off)
2020-01-04 05:02:17 -05:00
Brandon Thomas
9baea196d8
(not working, wip)
2020-01-04 04:33:47 -05:00
Brandon Thomas
25de25e3b0
Everything necessary to do scripting instead of tracing
2020-01-03 02:47:34 -05:00
Brandon Thomas
ad694e77a4
Change JIT mel tensor to float (from double). Works.
2020-01-03 02:35:44 -05:00
Brandon Thomas
ee5e48eae5
wip loading mels
2020-01-03 00:59:21 -05:00
Brandon Thomas
89e414225f
Fixes to get this working with JIT scripting (vs tracing)
2020-01-02 04:45:09 -05:00
Brandon Thomas
1ca7ee8a5b
rename models
2020-01-02 01:19:49 -05:00
Brandon Thomas
c0cac635f4
Add README to tacotron_melgan
2019-12-31 04:33:53 -05:00
Brandon Thomas
465a6f87f2
unsqueeze - still not working
2019-12-30 06:34:19 -05:00
Brandon Thomas
1f9235d56e
added text input, but it fails currently
2019-12-30 06:27:35 -05:00
Brandon Thomas
56e553ae66
Serving over Rocket HTTP works (very inelegant ATM)
2019-12-30 05:59:27 -05:00
Brandon Thomas
6c4d97e860
minor cleanup
2019-12-30 04:45:10 -05:00
Brandon Thomas
f51a7cd621
This works end-to-end
2019-12-30 04:15:34 -05:00
Brandon Thomas
89d08a449e
Rust Tacotron2 model runs, albeit SLOWLY
2019-12-30 03:48:24 -05:00
Brandon Thomas
4079ec379c
Tacotron model JIT works?
2019-12-30 03:41:57 -05:00
Brandon Thomas
3aacd5412e
add sequence file
2019-12-30 03:13:28 -05:00
Brandon Thomas
c986184fbf
This may have done it! (Python 3.6 upgrade helped)
2019-12-30 01:32:01 -05:00
Brandon Thomas
6985075187
This compiles but is garbage
...
- I removed Tacotron2.forward
2019-12-29 02:50:19 -05:00
Brandon Thomas
c86eeea550
minor progress, Tacotron JIT still does not work
2019-12-27 03:13:47 -05:00
Brandon Thomas
03cd0dc0db
model jit wrapping: got rid of postnet (wip)
2019-12-27 03:04:49 -05:00
Brandon Thomas
e492e2e10c
Added hparams and chopped off a few model pieces. Doesn't work yet
2019-12-27 02:46:05 -05:00
Brandon Thomas
2c33358db7
add nvidia's tacotron2 model
2019-12-27 02:31:30 -05:00
Brandon Thomas
546e5d7a09
Adding Tacotron2 wrapper (WIP)
2019-12-27 02:29:09 -05:00
Brandon Thomas
6982e65e0c
rename script
2019-12-27 00:24:18 -05:00
Brandon Thomas
c2b03c461a
seems pretty fast
2019-12-26 06:49:03 -05:00
Brandon Thomas
f051ab76a0
Now there's human voice audio!!!! WOO!
2019-12-26 06:44:40 -05:00
Brandon Thomas
57cbe02d7d
rename the model nodes
2019-12-26 06:33:42 -05:00
Brandon Thomas
60763c5120
I think this fixes the ResBlock
2019-12-26 06:01:00 -05:00
Brandon Thomas
3de3a3770c
add random sampling
2019-12-26 05:10:08 -05:00
Brandon Thomas
f07a628469
this makes clicking noises, but still mostly zero
2019-12-26 05:02:57 -05:00
Brandon Thomas
55947770cd
Model outputs audio, but all zeroes right now
2019-12-26 03:39:16 -05:00
Brandon Thomas
346855955c
Useful hacky script for jit containerizing mels/tensors
...
This seems to be the only(?) way of loading single tensors into
libtorch/tch.rs. It's a hack: load the tensor, pack it into a container
module, then define a forward method that returns the wrapped tensor.
2019-12-26 02:02:25 -05:00
Brandon Thomas
b9efb9db60
This might be a new approach...
2019-12-26 01:21:41 -05:00
Brandon Thomas
99d13ed283
light refactor
2019-12-25 17:01:28 -05:00
Brandon Thomas
615595c0d3
Add some mels for testing
2019-12-25 16:49:01 -05:00
Brandon Thomas
194c220ad2
Move melgan scripts to sub-project
2019-12-25 16:32:43 -05:00
Brandon Thomas
282e05340f
Now the two Tacotron projects work with the Rust Workspace
2019-12-25 01:54:02 -05:00
Brandon Thomas
a03966c209
Refactor into "workspace" project. VC sidecar works again!
...
Tensorflow Tacotron and Torch Tacotron+Melgan are broken (the latter
hasn't worked yet anyhow)
2019-12-25 01:31:19 -05:00