fail to pretrain

[chuanfushen]

Hi, Thanks for your great work!
When I train THuman using the script:

python train.py expname=pretrain

I obtain the below error.

Traceback (most recent call last):
  File "pretrain/train.py", line 124, in <module>
    main()
  File "/lab01/micromamba/envs/3dreid/lib/python3.8/site-packages/hydra/main.py", line 94, in decorated_main
    _run_hydra(
  File "/lab01/micromamba/envs/3dreid/lib/python3.8/site-packages/hydra/_internal/utils.py", line 394, in _run_hydra
    _run_app(
  File "/lab01/micromamba/envs/3dreid/lib/python3.8/site-packages/hydra/_internal/utils.py", line 457, in _run_app
    run_and_report(
  File "/lab01/micromamba/envs/3dreid/lib/python3.8/site-packages/hydra/_internal/utils.py", line 223, in run_and_report
    raise ex
  File "/lab01/micromamba/envs/3dreid/lib/python3.8/site-packages/hydra/_internal/utils.py", line 220, in run_and_report
    return func()
  File "/lab01/micromamba/envs/3dreid/lib/python3.8/site-packages/hydra/_internal/utils.py", line 458, in <lambda>
    lambda: hydra.run(
  File "/lab01/micromamba/envs/3dreid/lib/python3.8/site-packages/hydra/_internal/hydra.py", line 105, in run
    cfg = self.compose_config(
  File "/lab01/micromamba/envs/3dreid/lib/python3.8/site-packages/hydra/_internal/hydra.py", line 594, in compose_config
    cfg = self.config_loader.load_configuration(
  File "/lab01/micromamba/envs/3dreid/lib/python3.8/site-packages/hydra/_internal/config_loader_impl.py", line 142, in load_configuration
    return self._load_configuration_impl(
  File "/lab01/micromamba/envs/3dreid/lib/python3.8/site-packages/hydra/_internal/config_loader_impl.py", line 253, in _load_configuration_impl
    defaults_list = create_defaults_list(
  File "/lab01/micromamba/envs/3dreid/lib/python3.8/site-packages/hydra/_internal/defaults_list.py", line 745, in create_defaults_list
    defaults, tree = _create_defaults_list(
  File "/lab01/micromamba/envs/3dreid/lib/python3.8/site-packages/hydra/_internal/defaults_list.py", line 715, in _create_defaults_list
    defaults_tree = _create_defaults_tree(
  File "/lab01/micromamba/envs/3dreid/lib/python3.8/site-packages/hydra/_internal/defaults_list.py", line 356, in _create_defaults_tree
    ret = _create_defaults_tree_impl(
  File "/lab01/micromamba/envs/3dreid/lib/python3.8/site-packages/hydra/_internal/defaults_list.py", line 457, in _create_defaults_tree_impl
    return _expand_virtual_root(repo, root, overrides, skip_missing)
  File "/lab01/micromamba/envs/3dreid/lib/python3.8/site-packages/hydra/_internal/defaults_list.py", line 280, in _expand_virtual_root
    subtree = _create_defaults_tree_impl(
  File "/lab01/micromamba/envs/3dreid/lib/python3.8/site-packages/hydra/_internal/defaults_list.py", line 573, in _create_defaults_tree_impl
    add_child(children, new_root)
  File "/lab01/micromamba/envs/3dreid/lib/python3.8/site-packages/hydra/_internal/defaults_list.py", line 520, in add_child
    subtree_ = _create_defaults_tree_impl(
  File "/lab01/micromamba/envs/3dreid/lib/python3.8/site-packages/hydra/_internal/defaults_list.py", line 472, in _create_defaults_tree_impl
    overrides.set_known_choice(parent)
  File "/lab01/micromamba/envs/3dreid/lib/python3.8/site-packages/hydra/_internal/defaults_list.py", line 185, in set_known_choice
    raise ConfigCompositionException(
hydra.errors.ConfigCompositionException: Multiple values for network. To override a value use 'override network: texture'

[chuanfushen]

I have deal with this issue by modifying config.yaml:

defaults:
  - datamodule: data
hydra:
  run:
    dir: outputs/${expname}/

expname: joint
resume: false
starting_path: 'checkpoints/0.pth'
eval_mode: interp
seed: 42
agent_tot: 1
agent_id: 0

trainer:
  max_epochs: 2000
  gradient_clip_val: 0.1
  check_val_every_n_epoch: 10
  lr_scheduler_stepsize: [500,1000,1500]
  lr_schedule_decay_rate: 0.1
  deterministic: true
  gpus: -1
  log_every_n_steps: 5
  accelerator: 'ddp'
  profiler: 'simple'


model:
  dim_naked_shape: 512
  dim_clothed_shape: 512
  dim_texture: 512

  deformer:
    _target_: lib.model.deformer.ForwardDeformer
    max_steps: 50
    disp_network:
      _target_: lib.model.network.ImplicitNetwork
      d_in: 3
      d_out: 3
      width: 128
      depth: 4
      geometric_init: false
      bias: 1
      weight_norm: true
      multires: 0
      pose_cond_dim: 69
      pose_cond_layer: []
      pose_embed_dim: -1
      shape_cond_dim: 512
      shape_cond_layer: [0]
      shape_embed_dim: -1
      latent_cond_dim: 128
      latent_cond_layer: []
      latent_embed_dim: -1
      final_acti_type: none
    lbs_network:
      _target_: lib.model.network.ImplicitNetwork
      d_in: 3
      d_out: 24
      width: 128
      depth: 4
      geometric_init: false
      bias: 1
      weight_norm: true
      multires: 0
      pose_cond_dim: 69
      pose_cond_layer: []
      pose_embed_dim: -1
      shape_cond_dim: 10
      shape_cond_layer: []
      shape_embed_dim: -1
      latent_cond_dim: 512
      latent_cond_layer: [0]
      latent_embed_dim: -1
      final_acti_type: none

  clothed_network:
    _target_: lib.model.network.ImplicitNetwork
    d_in: 3
    d_out: 257
    width: 512
    depth: 8
    geometric_init: true
    bias: 1
    skip_in: [4]
    weight_norm: true
    multires: 6
    pose_cond_dim: 69
    pose_cond_layer: []
    pose_embed_dim: -1
    shape_cond_dim: 10
    shape_cond_layer: []
    shape_embed_dim: -1
    latent_cond_dim: 512
    latent_cond_layer: [0]
    latent_embed_dim: -1
    feat_cond_dim: 256
    feat_cond_layer: [0]
    final_acti_type: none

  naked_network:
    _target_: lib.model.network.ImplicitNetwork
    d_in: 3
    d_out: 257
    width: 512
    depth: 8 #4
    geometric_init: true
    bias: 1
    skip_in: [4]
    weight_norm: true
    multires: 6
    pose_cond_dim: 69
    pose_cond_layer: []
    pose_embed_dim: -1
    shape_cond_dim: 10
    shape_cond_layer: []
    shape_embed_dim: -1
    latent_cond_dim: 64
    latent_cond_layer: [0]
    latent_embed_dim: -1
    final_acti_type: none

  texture_network:
    _target_: lib.model.network.ImplicitNetwork
    d_in: 3
    d_out: 3
    width: 256 #128
    depth: 6 #4
    weight_norm: true
    multires: 6
    pose_cond_dim: 69
    pose_cond_layer: []
    pose_embed_dim: -1
    shape_cond_dim: 10
    shape_cond_layer: []
    shape_embed_dim: -1
    latent_cond_dim: 512
    latent_cond_layer: [0, 4]
    skip_in: []
    latent_embed_dim: -1
    feat_cond_dim: 256
    feat_cond_layer: [0]


  optim:
    lr: 1e-3

  nepochs_pretrain: 10    #1
  pretrain_bone: true
  nepochs_pretrain_coarse: 0    #1
  lambda_bone_occ: 1
  lambda_bone_w: 10
  lambda_disp: 10
  lambda_reg: 1e-3

[Bruce-GuoS]

I have deal with this issue by modifying config.yaml:

defaults:
  - datamodule: data
hydra:
  run:
    dir: outputs/${expname}/

expname: joint
resume: false
starting_path: 'checkpoints/0.pth'
eval_mode: interp
seed: 42
agent_tot: 1
agent_id: 0

trainer:
  max_epochs: 2000
  gradient_clip_val: 0.1
  check_val_every_n_epoch: 10
  lr_scheduler_stepsize: [500,1000,1500]
  lr_schedule_decay_rate: 0.1
  deterministic: true
  gpus: -1
  log_every_n_steps: 5
  accelerator: 'ddp'
  profiler: 'simple'


model:
  dim_naked_shape: 512
  dim_clothed_shape: 512
  dim_texture: 512

  deformer:
    _target_: lib.model.deformer.ForwardDeformer
    max_steps: 50
    disp_network:
      _target_: lib.model.network.ImplicitNetwork
      d_in: 3
      d_out: 3
      width: 128
      depth: 4
      geometric_init: false
      bias: 1
      weight_norm: true
      multires: 0
      pose_cond_dim: 69
      pose_cond_layer: []
      pose_embed_dim: -1
      shape_cond_dim: 512
      shape_cond_layer: [0]
      shape_embed_dim: -1
      latent_cond_dim: 128
      latent_cond_layer: []
      latent_embed_dim: -1
      final_acti_type: none
    lbs_network:
      _target_: lib.model.network.ImplicitNetwork
      d_in: 3
      d_out: 24
      width: 128
      depth: 4
      geometric_init: false
      bias: 1
      weight_norm: true
      multires: 0
      pose_cond_dim: 69
      pose_cond_layer: []
      pose_embed_dim: -1
      shape_cond_dim: 10
      shape_cond_layer: []
      shape_embed_dim: -1
      latent_cond_dim: 512
      latent_cond_layer: [0]
      latent_embed_dim: -1
      final_acti_type: none

  clothed_network:
    _target_: lib.model.network.ImplicitNetwork
    d_in: 3
    d_out: 257
    width: 512
    depth: 8
    geometric_init: true
    bias: 1
    skip_in: [4]
    weight_norm: true
    multires: 6
    pose_cond_dim: 69
    pose_cond_layer: []
    pose_embed_dim: -1
    shape_cond_dim: 10
    shape_cond_layer: []
    shape_embed_dim: -1
    latent_cond_dim: 512
    latent_cond_layer: [0]
    latent_embed_dim: -1
    feat_cond_dim: 256
    feat_cond_layer: [0]
    final_acti_type: none

  naked_network:
    _target_: lib.model.network.ImplicitNetwork
    d_in: 3
    d_out: 257
    width: 512
    depth: 8 #4
    geometric_init: true
    bias: 1
    skip_in: [4]
    weight_norm: true
    multires: 6
    pose_cond_dim: 69
    pose_cond_layer: []
    pose_embed_dim: -1
    shape_cond_dim: 10
    shape_cond_layer: []
    shape_embed_dim: -1
    latent_cond_dim: 64
    latent_cond_layer: [0]
    latent_embed_dim: -1
    final_acti_type: none

  texture_network:
    _target_: lib.model.network.ImplicitNetwork
    d_in: 3
    d_out: 3
    width: 256 #128
    depth: 6 #4
    weight_norm: true
    multires: 6
    pose_cond_dim: 69
    pose_cond_layer: []
    pose_embed_dim: -1
    shape_cond_dim: 10
    shape_cond_layer: []
    shape_embed_dim: -1
    latent_cond_dim: 512
    latent_cond_layer: [0, 4]
    skip_in: []
    latent_embed_dim: -1
    feat_cond_dim: 256
    feat_cond_layer: [0]


  optim:
    lr: 1e-3

  nepochs_pretrain: 10    #1
  pretrain_bone: true
  nepochs_pretrain_coarse: 0    #1
  lambda_bone_occ: 1
  lambda_bone_w: 10
  lambda_disp: 10
  lambda_reg: 1e-3

Hello, did you reproduce the results of the author's?

[chuanfushen]

not really, I am currently working on other project, so I don't continue working on this repo anymore