forked from mrq/ai-voice-cloning
19 lines
518 B
YAML
19 lines
518 B
YAML
|
path:
|
||
|
|
pretrain_model_dvae: './models/tortoise/dvae.pth'
|
||
|
|
strict_load: true
|
||
|
|
#resume_state: ../experiments/train_imgnet_vqvae_stage1/training_state/0.state # <-- Set this to resume from a previous training state.
|
||
|
|
networks:
|
||
|
|
dvae:
|
||
|
|
type: generator
|
||
|
|
which_model_G: lucidrains_dvae
|
||
|
|
kwargs:
|
||
|
|
channels: 80
|
||
|
|
codebook_dim: 512
|
||
|
|
hidden_dim: 512
|
||
|
|
kernel_size: 3
|
||
|
|
num_layers: 2
|
||
|
|
num_resnet_blocks: 3
|
||
|
|
num_tokens: 8192
|
||
|
|
positional_dims: 1
|
||
|
|
use_transposed_convs: false
|