# Experiment general info
name: "OV_ScanQA"
rng_seed: 1234
num_gpu: 1
mode: "train"
note: ""
# Choose keywords to feature your saving directory
naming_keywords: ["dataloader.batchsize", "task", "note", "time"]
base_dir: "results"
exp_dir: ""
save_frequency: 100

resume: False

debug:
  flag: False
  hard_debug: False
  debug_size: 20

logger:
  name: "wandb"
  entity: "yem"

# dataset details
data:
  train: ['ScanNetScanQAOld']
  val: ['ScanNetScanQAOld']
  test: ['ScanNetScanQAOld']
  args:
    max_obj_len: 80
    max_seq_len: 50
    num_points: 1024
    pc_type: 'pred'
    sem_type: '607'
    filter_lang: False
    rot_aug: True
  ScanNetScanQAOld:
    train:
      use_unanswer: True
    val:
      use_unanswer: True
    test:
      use_unanswer: True
      test_file: "test_w_obj" # or "test_wo_obj"

  use_voxel: False
  scan_family_base: "PointMapVerse/existing_datasets/ScanNet"
  rscan_base: "PointMapVerse/existing_datasets/3RScan"

# task details: 'pretrain', ' scanrefer', 'referit3d', 'scanqa', 'default'
task: 'ScanQA'
data_wrapper:
  train: 'ScanFamilyDatasetWrapperQA'
  val: 'ScanFamilyDatasetWrapperQA'
  test: 'ScanFamilyDatasetWrapperQA'

# Training details
trainer: "DefaultTrainer"
ckpt_path: ""
pretrain_ckpt_path: "/home/m50048399/transfered/ye_project/UniPointMap/results/scanqa_sft_run1_b64_ScanQA_ScanNetScanQAOld_scanqa_sft_run1/2025-10-08-14:59:55.120964/ckpt/best.pth"

# dataloader details
dataloader:
  # This is a per-gpu batchsize
  batchsize: 64
  num_workers: 2
  balance_dataset: False
  filter_empty_annotations: False

solver:
  gradient_accumulation_steps: 1
  epochs_per_save: 100
  epochs_per_eval: 1
  lr: 1e-4
  grad_norm: 5.0
  epochs: 100
  optim:
    name: "AdamW"
    args:
      betas: [0.9, 0.98]
  sched:
    name: "warmup_cosine"
    args:
      warmup_steps: 5000

eval:
  name: "ScanQAEval"
  save: False

# Model details
model:
  name: OpenVocab
  language:
    # This part could be further optimized to be using
    # huggingface yaml config files
    name: "BERTLanguageEncoder"
    args:
      weights: "fg-clip-base"
      hidden_size: 768
      num_hidden_layers: 4
      num_attention_heads: 12
      type_vocab_size: 2
    lr: 1e-4
  vision:
#    name: "pointnet_point_encoder"
#    args:
#      path: None
#      freeze: False
    name: 'fg-clip-base'
    args:
        backbone: "pointnet++"
        hidden_size: 768
        freeze: True
        path: 'fg-clip-base'
        num_attention_heads: 12
        spatial_dim: 5
        num_layers: 4
        dim_loc: 6
        dim_feedforward: 2048
        attn_type: spatial
        pairwise_rel_type: 'center'
        use_matmul_label: False
        lang_type: 'bert'
        lang_path: 'pretrained_weights/607_text_embeddings'
    lr: 1e-4
  grounding:
    name: 'UnifiedSpatialCrossEncoderV2'
    args:
      hidden_size: 1024
      num_attention_heads: 16
      num_layers: 4
      dim_feedforward: 2048
      dim_loc: 6
    lr: 1e-4
  inter: before
  heads:
    head_list: ["qa_head"]
    qa_head:
      name: "QAHeadV1"
      args:
        hidden_size: 768
        mlp_size: 256
        glimpse: 1
        flat_out_size: 512
        num_answers: 8864
  loss_type: "ListLoss"
  loss_list: [
      "answer_loss",
      # 'TextObjWithinBatch',
  ]
  vis_loss_list: [
      "answer_loss",
      # 'TextObjWithinBatch',
  ]