Upload 5 files
Browse files- checkpoints/240112_icl_audio2secc_vox2_cmlr/config.yaml +86 -0
- checkpoints/240112_icl_audio2secc_vox2_cmlr/model_ckpt_steps_1856000.ckpt +3 -0
- checkpoints/mimictalk_orig/os_secc2plane_torso/config.yaml +152 -0
- checkpoints/mimictalk_orig/os_secc2plane_torso/model_ckpt_steps_100000.ckpt +3 -0
- checkpoints/pretrained_ckpts/mit_b0.pth +3 -0
checkpoints/240112_icl_audio2secc_vox2_cmlr/config.yaml
ADDED
|
@@ -0,0 +1,86 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
accumulate_grad_batches: 1
|
| 2 |
+
amp: false
|
| 3 |
+
audio_type: hubert
|
| 4 |
+
base_config:
|
| 5 |
+
- ./audio2secc_vae.yaml
|
| 6 |
+
batch_size: 4
|
| 7 |
+
binarization_args:
|
| 8 |
+
with_coeff: true
|
| 9 |
+
with_hubert: true
|
| 10 |
+
with_mel: true
|
| 11 |
+
binary_data_dir: data/binary/voxceleb2_audio2motion
|
| 12 |
+
blink_mode: blink_unit
|
| 13 |
+
clip_grad_norm: 0.5
|
| 14 |
+
clip_grad_value: 0
|
| 15 |
+
debug: false
|
| 16 |
+
ds_name: Concat_voxceleb2_CMLR
|
| 17 |
+
eval_max_batches: 10
|
| 18 |
+
gen_dir_name: ''
|
| 19 |
+
hidden_size: 256
|
| 20 |
+
icl_model_type: icl_flow_matching
|
| 21 |
+
infer_audio_source_name: ''
|
| 22 |
+
infer_ckpt_steps: 40000
|
| 23 |
+
infer_out_npy_name: ''
|
| 24 |
+
init_from_ckpt: ''
|
| 25 |
+
lambda_kl: 0.02
|
| 26 |
+
lambda_kl_t1: 2000
|
| 27 |
+
lambda_kl_t2: 2000
|
| 28 |
+
lambda_l2_reg_exp: 0.0
|
| 29 |
+
lambda_lap_exp: 0.0
|
| 30 |
+
lambda_lap_exp_x1: 0.1
|
| 31 |
+
lambda_mse_exp: 0.0
|
| 32 |
+
lambda_mse_exp_x1: 0.1
|
| 33 |
+
lambda_mse_icl: 1.0
|
| 34 |
+
lambda_mse_lm2d: 0.0
|
| 35 |
+
lambda_mse_lm3d: 0.0
|
| 36 |
+
lambda_mse_lm3d_x1: 0.1
|
| 37 |
+
lambda_sync_lm3d: 0.05
|
| 38 |
+
load_ckpt: ''
|
| 39 |
+
load_db_to_memory: false
|
| 40 |
+
lr: 0.0005
|
| 41 |
+
max_sentences_per_batch: 512
|
| 42 |
+
max_tokens_per_batch: 20000
|
| 43 |
+
max_updates: 4000000
|
| 44 |
+
motion_type: exp
|
| 45 |
+
num_ckpt_keep: 100
|
| 46 |
+
num_sanity_val_steps: 5
|
| 47 |
+
num_valid_plots: 1
|
| 48 |
+
num_workers: 8
|
| 49 |
+
optimizer_adam_beta1: 0.9
|
| 50 |
+
optimizer_adam_beta2: 0.999
|
| 51 |
+
print_nan_grads: false
|
| 52 |
+
process_id: 0
|
| 53 |
+
raw_data_dir: /home/tiger/datasets/raw/TH1KH_512
|
| 54 |
+
ref_id_mode: first_frame
|
| 55 |
+
resume_from_checkpoint: 0
|
| 56 |
+
sample_min_length: 32
|
| 57 |
+
save_best: false
|
| 58 |
+
save_codes:
|
| 59 |
+
- tasks
|
| 60 |
+
- modules
|
| 61 |
+
- egs
|
| 62 |
+
save_gt: true
|
| 63 |
+
scheduler: exponential
|
| 64 |
+
seed: 9999
|
| 65 |
+
smo_win_size: 5
|
| 66 |
+
split_seed: 999
|
| 67 |
+
syncnet_ckpt_dir: checkpoints/0904_syncnet/syncnet_hubert_vox2
|
| 68 |
+
task_cls: tasks.os_avatar.icl_audio2secc_task.Audio2SECCTask
|
| 69 |
+
tb_log_interval: 100
|
| 70 |
+
total_process: 1
|
| 71 |
+
use_aux_features: true
|
| 72 |
+
use_aux_loss_on_x1: true
|
| 73 |
+
use_eye_amp_embed: false
|
| 74 |
+
use_flow: true
|
| 75 |
+
use_fork: true
|
| 76 |
+
use_kv_dataset: true
|
| 77 |
+
use_mouth_amp_embed: true
|
| 78 |
+
use_pitch: true
|
| 79 |
+
val_check_interval: 2000
|
| 80 |
+
valid_infer_interval: 2000
|
| 81 |
+
valid_monitor_key: val_loss
|
| 82 |
+
valid_monitor_mode: min
|
| 83 |
+
warmup_updates: 1000
|
| 84 |
+
weight_decay: 0
|
| 85 |
+
work_dir: checkpoints/240112_audio2secc/icl_audio2secc_vox2_cmlr
|
| 86 |
+
x_multiply: 16
|
checkpoints/240112_icl_audio2secc_vox2_cmlr/model_ckpt_steps_1856000.ckpt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:367167db3b25fe07de9255871ce7813158551c4b00bfadd7334a671648924a2e
|
| 3 |
+
size 462941609
|
checkpoints/mimictalk_orig/os_secc2plane_torso/config.yaml
ADDED
|
@@ -0,0 +1,152 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
accumulate_grad_batches: 1
|
| 2 |
+
add_ffhq_singe_disc: false
|
| 3 |
+
also_update_decoder: false
|
| 4 |
+
amp: false
|
| 5 |
+
base_channel: 32768
|
| 6 |
+
base_config:
|
| 7 |
+
- ./secc_img2plane.yaml
|
| 8 |
+
batch_size: 1
|
| 9 |
+
binary_data_dir: data/binary/CelebV-HQ
|
| 10 |
+
blur_fade_kimg: 20
|
| 11 |
+
blur_init_sigma: 10
|
| 12 |
+
blur_raw_target: true
|
| 13 |
+
box_warp: 1
|
| 14 |
+
ckpt_milestone_interval: 50000
|
| 15 |
+
clip_grad_norm: 1.0
|
| 16 |
+
clip_grad_value: 0
|
| 17 |
+
cond_hid_dim: 32
|
| 18 |
+
cond_out_dim: 16
|
| 19 |
+
cond_type: idexp_lm3d_normalized
|
| 20 |
+
debug: false
|
| 21 |
+
density_reg_p_dist: 0.004
|
| 22 |
+
disable_highreso_at_stage1: true
|
| 23 |
+
disc_c_noise: 1.0
|
| 24 |
+
disc_cond_mode: none
|
| 25 |
+
ds_name: Concat_VFHQ_CelebVHQ_TH1KH_RAVDESS
|
| 26 |
+
ema_interval: 400
|
| 27 |
+
enable_rescale_plane_regulation: false
|
| 28 |
+
eval_max_batches: 100
|
| 29 |
+
ffhq_disc_inp_mode: eg3d_gen
|
| 30 |
+
final_resolution: 512
|
| 31 |
+
flipped_to_world_coord: true
|
| 32 |
+
fuse_with_deform_source: false
|
| 33 |
+
gen_cond_mode: none
|
| 34 |
+
generator_condition_on_pose: true
|
| 35 |
+
gpc_reg_fade_kimg: 1000
|
| 36 |
+
gpc_reg_prob: 0.8
|
| 37 |
+
group_size_for_mini_batch_std: 2
|
| 38 |
+
htbsr_head_threshold: 1.0
|
| 39 |
+
htbsr_head_weight_fuse_mode: v2
|
| 40 |
+
img2plane_backbone_mode: composite
|
| 41 |
+
img2plane_backbone_scale: standard
|
| 42 |
+
init_from_ckpt: checkpoints/240118_os_secc2planes/os_secc2plane_pertubeBlink0.05_pertubeSECC0.05
|
| 43 |
+
lam_occlusion_2_reg_l1: 0.0
|
| 44 |
+
lam_occlusion_reg_l1: 0.0
|
| 45 |
+
lam_occlusion_weights_entropy: 0.001
|
| 46 |
+
lambda_G_adversarial_adv: 1.0
|
| 47 |
+
lambda_G_supervise_adv: 1.0
|
| 48 |
+
lambda_G_supervise_mse: 1.0
|
| 49 |
+
lambda_G_supervise_mse_raw: 1.0
|
| 50 |
+
lambda_density_reg: 0.25
|
| 51 |
+
lambda_ffhq_mv_adv: 0.002
|
| 52 |
+
lambda_gradient_penalty: 1.0
|
| 53 |
+
lambda_mse: 1.0
|
| 54 |
+
lambda_mse_depth: 0.0
|
| 55 |
+
lambda_th1kh_mv_adv: 0.003
|
| 56 |
+
lambda_weights_entropy: 0.01
|
| 57 |
+
lambda_weights_l1: 0.1
|
| 58 |
+
load_ckpt: ''
|
| 59 |
+
lpips_mode: vgg19_v2
|
| 60 |
+
lr_d: 0.0002
|
| 61 |
+
lr_decay_interval: 5000
|
| 62 |
+
lr_decay_rate: 0.95
|
| 63 |
+
lr_g: 1.0e-05
|
| 64 |
+
lr_lambda_pertube_secc: 0.01
|
| 65 |
+
lr_mul_cano_img2plane: 1.0
|
| 66 |
+
mapping_network_depth: 2
|
| 67 |
+
max_channel: 512
|
| 68 |
+
max_updates: 100000
|
| 69 |
+
mimic_plane: false
|
| 70 |
+
min_rescale_factor: 0.25
|
| 71 |
+
motion_smo_win_size: 5
|
| 72 |
+
neural_rendering_resolution: 128
|
| 73 |
+
normalize_cond: false
|
| 74 |
+
normalize_radius: false
|
| 75 |
+
not_save_modules:
|
| 76 |
+
- criterion_lpips
|
| 77 |
+
- eg3d_model
|
| 78 |
+
num_ckpt_keep: 1
|
| 79 |
+
num_fp16_layers_in_discriminator: 4
|
| 80 |
+
num_fp16_layers_in_generator: 0
|
| 81 |
+
num_fp16_layers_in_super_resolution: 4
|
| 82 |
+
num_samples_coarse: 48
|
| 83 |
+
num_samples_fine: 48
|
| 84 |
+
num_sanity_val_steps: 1
|
| 85 |
+
num_valid_plots: 25
|
| 86 |
+
num_workers: 8
|
| 87 |
+
occlusion_fuse: true
|
| 88 |
+
ones_ws_for_sr: true
|
| 89 |
+
optimizer_adam_beta1_d: 0.0
|
| 90 |
+
optimizer_adam_beta1_g: 0.0
|
| 91 |
+
optimizer_adam_beta2_d: 0.99
|
| 92 |
+
optimizer_adam_beta2_g: 0.99
|
| 93 |
+
phase1_plane_fusion_mode: add
|
| 94 |
+
pncc_cond_mode: cano_src_tgt
|
| 95 |
+
pretrained_eg3d_ckpt: /mnt/bn/ailabrenyi/entries/yezhenhui/projects/GeneFace_private/checkpoints/0628_img2planes/eg3d_baseline_run2/model_ckpt_steps_100000.ckpt
|
| 96 |
+
print_nan_grads: false
|
| 97 |
+
process_id: 0
|
| 98 |
+
processed_data_dir: data/processed/videos
|
| 99 |
+
random_sample_pose: true
|
| 100 |
+
raw_data_dir: /home/tiger/datasets/raw/FFHQ
|
| 101 |
+
ray_far: auto
|
| 102 |
+
ray_near: auto
|
| 103 |
+
reg_interval_d: 16
|
| 104 |
+
reg_interval_g: 4
|
| 105 |
+
reg_interval_g_cond: 4
|
| 106 |
+
reload_head_ckpt: ''
|
| 107 |
+
resume_from_checkpoint: 0
|
| 108 |
+
save_best: true
|
| 109 |
+
save_codes:
|
| 110 |
+
- tasks
|
| 111 |
+
- modules
|
| 112 |
+
- egs
|
| 113 |
+
secc_pertube_mode: randn
|
| 114 |
+
secc_pertube_randn_scale: 0.01
|
| 115 |
+
secc_segformer_scale: b0
|
| 116 |
+
seed: 9999
|
| 117 |
+
seg_out_mode: head
|
| 118 |
+
smo_win_size: 5
|
| 119 |
+
split_seed: 999
|
| 120 |
+
sr_type: vanilla
|
| 121 |
+
start_adv_iters: 40000
|
| 122 |
+
target_pertube_blink_secc_loss: 0.05
|
| 123 |
+
target_pertube_secc_loss: 0.05
|
| 124 |
+
task_cls: tasks.os_avatar.secc_img2plane_torso_task.SECC_Img2PlaneEG3D_TorsoTask
|
| 125 |
+
tb_log_interval: 100
|
| 126 |
+
torch_compile: true
|
| 127 |
+
torso_kp_num: 4
|
| 128 |
+
torso_model_version: v2
|
| 129 |
+
torso_occlusion_reg_unmask_factor: 0.3
|
| 130 |
+
torso_ref_segout_mode: torso
|
| 131 |
+
total_process: 1
|
| 132 |
+
triplane_depth: 3
|
| 133 |
+
triplane_feature_type: trigrid_v2
|
| 134 |
+
triplane_hid_dim: 32
|
| 135 |
+
two_stage_training: true
|
| 136 |
+
update_on_th1kh_samples: false
|
| 137 |
+
update_src2src_interval: 4
|
| 138 |
+
use_kv_dataset: true
|
| 139 |
+
use_motion_smo_net: false
|
| 140 |
+
use_mse: false
|
| 141 |
+
use_th1kh_disc: false
|
| 142 |
+
use_th1kh_mv_adv: false
|
| 143 |
+
val_check_interval: 2000
|
| 144 |
+
valid_infer_interval: 2000
|
| 145 |
+
valid_monitor_key: val_loss
|
| 146 |
+
valid_monitor_mode: min
|
| 147 |
+
video_id: May
|
| 148 |
+
w_dim: 512
|
| 149 |
+
warmup_updates: 4000
|
| 150 |
+
weight_fuse: true
|
| 151 |
+
work_dir: checkpoints/240120_os_secc2planes_torso/os_secc2plane_torso_htbsrFusev2_htbsrThres1.0
|
| 152 |
+
z_dim: 512
|
checkpoints/mimictalk_orig/os_secc2plane_torso/model_ckpt_steps_100000.ckpt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:757c1b73d938da0ec3fba555a155a31ac803ddc8d343bba2a3c69845844b213a
|
| 3 |
+
size 1414788463
|
checkpoints/pretrained_ckpts/mit_b0.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:df468f7f13c4186f25bd3e2caf09e4f927b5b5ac0abccac84011dae747d4c49c
|
| 3 |
+
size 14331578
|