Commit
·
9d604b4
1
Parent(s):
afbdfd0
Upload 2 files
Browse files- diffusion.yaml +75 -0
- model_440000.pt +3 -0
diffusion.yaml
ADDED
|
@@ -0,0 +1,75 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
data:
|
| 2 |
+
block_size: 512
|
| 3 |
+
cnhubertsoft_gate: 10
|
| 4 |
+
duration: 1
|
| 5 |
+
encoder: vec256l9
|
| 6 |
+
encoder_hop_size: 320
|
| 7 |
+
encoder_out_channels: 256
|
| 8 |
+
encoder_sample_rate: 16000
|
| 9 |
+
extensions:
|
| 10 |
+
- wav
|
| 11 |
+
sampling_rate: 44100
|
| 12 |
+
training_files: filelists/train.txt
|
| 13 |
+
validation_files: filelists/val.txt
|
| 14 |
+
device: cuda
|
| 15 |
+
env:
|
| 16 |
+
expdir: logs/44k/diffusion
|
| 17 |
+
gpu_id: 0
|
| 18 |
+
infer:
|
| 19 |
+
method: dpm-solver
|
| 20 |
+
speedup: 10
|
| 21 |
+
model:
|
| 22 |
+
n_chans: 512
|
| 23 |
+
n_hidden: 256
|
| 24 |
+
n_layers: 20
|
| 25 |
+
n_spk: 28
|
| 26 |
+
type: Diffusion
|
| 27 |
+
use_pitch_aug: true
|
| 28 |
+
spk:
|
| 29 |
+
biaobei: 0
|
| 30 |
+
buyizi: 1
|
| 31 |
+
kirizu: 2
|
| 32 |
+
m4_alto_1: 3
|
| 33 |
+
m4_alto_2: 4
|
| 34 |
+
m4_alto_3: 5
|
| 35 |
+
m4_alto_4: 6
|
| 36 |
+
m4_alto_5: 7
|
| 37 |
+
m4_alto_6: 8
|
| 38 |
+
m4_alto_7: 9
|
| 39 |
+
m4_bass_1: 10
|
| 40 |
+
m4_bass_2: 11
|
| 41 |
+
m4_bass_3: 12
|
| 42 |
+
m4_soprano_1: 13
|
| 43 |
+
m4_soprano_2: 14
|
| 44 |
+
m4_soprano_3: 15
|
| 45 |
+
m4_tenor_1: 16
|
| 46 |
+
m4_tenor_2: 17
|
| 47 |
+
m4_tenor_3: 18
|
| 48 |
+
m4_tenor_4: 19
|
| 49 |
+
m4_tenor_5: 20
|
| 50 |
+
m4_tenor_6: 21
|
| 51 |
+
m4_tenor_7: 22
|
| 52 |
+
nyaru: 23
|
| 53 |
+
opencpop: 24
|
| 54 |
+
paimon: 25
|
| 55 |
+
sanyueqi: 26
|
| 56 |
+
yousa: 27
|
| 57 |
+
train:
|
| 58 |
+
amp_dtype: fp16
|
| 59 |
+
batch_size: 48
|
| 60 |
+
cache_all_data: false
|
| 61 |
+
cache_device: cpu
|
| 62 |
+
cache_fp16: true
|
| 63 |
+
decay_step: 100000
|
| 64 |
+
epochs: 100000
|
| 65 |
+
gamma: 0.5
|
| 66 |
+
interval_force_save: 10000
|
| 67 |
+
interval_log: 10
|
| 68 |
+
interval_val: 2000
|
| 69 |
+
lr: 0.0002
|
| 70 |
+
num_workers: 2
|
| 71 |
+
save_opt: false
|
| 72 |
+
weight_decay: 0
|
| 73 |
+
vocoder:
|
| 74 |
+
ckpt: pretrain/nsf_hifigan/model
|
| 75 |
+
type: nsf-hifigan
|
model_440000.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:134e457e60a933c41a18f159032737bbe3f3713c84217ea60dc800f04c4dfffe
|
| 3 |
+
size 220398667
|