Add files using upload-large-folder tool

light_scannet/scene0697_01.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2a8b855f07dacd32c54e96131869f36953d253bc728b12c732cac2366e1d7819
+size 169390352

light_scannet/scene0701_00.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:50d41ec159e4c74f981e036794b836af7b702dcc3e2584a3860958de427baf98
+size 169390352

light_scannet/scene0701_01.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:771a75137ae85585afbe0ff22a74e1f39872102d9a07f3ba23f39dc4636f8676
+size 169390352

light_scannet/scene0701_02.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:37c3fe835ab97759480360cc88525a063f6fa542c2a2effad6a73eb02aff7c5c
+size 169390352

light_scannet/scene0702_00.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:43872b9b1ea182de2a28a9ac2027008edeee893435019ef3eace55d7619554de
+size 169390352

light_scannet/scene0702_01.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8d8b2f1d63e088ded48e97a6aefe963c3e0f25a2991fa2039126cd28572923ec
+size 169390352

light_scannet/scene0702_02.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19a3728092db64ddce530719a05e4a399f6bbc337da04faeb43e331c06d8071b
+size 169390352

light_scannet/scene0703_00.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:34e28a92a0ac6381e973fc97e22fb9e3ecd19d5e88ffb16faec043a29a577bba
+size 169390352

light_scannet/scene0703_01.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dda357be633a2c224989ff8bc5822bb1215d140acff816e5814bc7474ee0a349
+size 169390352

light_scannet/scene0706_00.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9ed432423b6130fa8745077aa70b662538af7d942c010e834cb35c84991deb6e
+size 169390352

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-06-27-20:57:16.270199/ckpt/best.pth/model_1.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:60d95b10b6e140a9626a7058d5038528f2ff80148dc4569b881db56052046509
+size 40

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:17:42.163133/config.yaml

@@ -0,0 +1,139 @@
+name: sqa3d_sft_align_run1
+rng_seed: 42
+num_gpu: 4
+mode: train
+note: sqa3d_sft_align_run1
+naming_keywords:
+- dataloader.batchsize
+- task
+- note
+- time
+base_dir: results
+exp_dir: !!python/object/apply:pathlib.PosixPath
+- results
+- sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1
+- 2025-07-02-19:17:42.163133
+save_frequency: 10
+resume: false
+is_pretrain: false
+debug:
+  flag: false
+  hard_debug: false
+  debug_size: 20
+logger:
+  name: wandb
+  entity: yem
+  run_id: uogvgf63
+data:
+  train:
+  - ScanNetSQA3D
+  val:
+  - ScanNetSQA3D
+  test:
+  - ScanNetSQA3D
+  args:
+    max_obj_len: 80
+    max_seq_len: 50
+    num_points: 1024
+    pc_type: pred
+    sem_type: '607'
+    filter_lang: false
+    rot_aug: true
+  ScanNetSQA3D:
+    train:
+      use_unanswer: true
+    val:
+      use_unanswer: true
+    test:
+      use_unanswer: true
+  use_voxel: false
+  scan_family_base: existing_datasets/ScanNet
+  rscan_base: existing_datasets/3RScan
+task: SQA3D
+data_wrapper:
+  train: ScanFamilyDatasetWrapperOld
+  val: ScanFamilyDatasetWrapperOld
+  test: ScanFamilyDatasetWrapperOld
+trainer: DefaultTrainer
+ckpt_path: ''
+pretrain_ckpt_path: /home/m50048399/SceneVerse/results/sceneverse_scannet_exp1_b32_Pretrain_all_scannet_training_run1/b32_contrastive_mpm_random_caption/ckpt/ckpt_140.pth
+dataloader:
+  batchsize: 6
+  num_workers: 2
+  balance_dataset: false
+  filter_empty_annotations: false
+solver:
+  gradient_accumulation_steps: 1
+  epochs_per_save: 10
+  epochs_per_eval: 1
+  lr: 0.0001
+  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: SQA3DEval
+  save: false
+model:
+  name: OpenVocab
+  language:
+    name: BERTLanguageEncoder
+    args:
+      weights: bert-base-uncased
+      hidden_size: 768
+      num_hidden_layers: 4
+      num_attention_heads: 12
+      type_vocab_size: 2
+    lr: 1.0e-05
+  vision:
+    name: qihoo360/fg-clip-base
+    args:
+      backbone: pointnet++
+      hidden_size: 768
+      freeze: true
+      path: pretrained_weights/pointnetpp-open-bert
+      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: 0.0001
+  grounding:
+    name: UnifiedSpatialCrossEncoderV2
+    args:
+      hidden_size: 768
+      num_attention_heads: 12
+      num_layers: 4
+      dim_feedforward: 2048
+      dim_loc: 6
+    lr: 0.0001
+  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: 706
+  loss_type: ListLoss
+  loss_list:
+  - answer_loss
+  vis_loss_list:
+  - answer_loss

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:17:42.164767/ckpt/best.pth/custom_checkpoint_0.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0766a661767bcd76fe4c5da8bec73ff8ca2a8a1997ba80b17273e8f0ddf2bd3a
+size 1040

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:17:42.164767/ckpt/best.pth/model_1.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:60d95b10b6e140a9626a7058d5038528f2ff80148dc4569b881db56052046509
+size 40

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:17:42.164767/ckpt/best.pth/random_states_0.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e2d29f92a09daef7a7c2cc2fad629d0c329ba7c3bc3cd1c94a9bd32d0782d52
+size 15124

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:17:42.164767/ckpt/best.pth/scheduler.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a02ba49a043e4028cddd88477a825c75b63a511de04c5483cd6fe1191fbc986
+size 1064

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:17:42.164767/ckpt/latest.pth/custom_checkpoint_0.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0766a661767bcd76fe4c5da8bec73ff8ca2a8a1997ba80b17273e8f0ddf2bd3a
+size 1040

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:17:42.164767/ckpt/latest.pth/model_1.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:60d95b10b6e140a9626a7058d5038528f2ff80148dc4569b881db56052046509
+size 40

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:17:42.164767/ckpt/latest.pth/random_states_0.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e2d29f92a09daef7a7c2cc2fad629d0c329ba7c3bc3cd1c94a9bd32d0782d52
+size 15124

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:17:42.164767/ckpt/latest.pth/scheduler.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a02ba49a043e4028cddd88477a825c75b63a511de04c5483cd6fe1191fbc986
+size 1064

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:17:42.164767/config.yaml

@@ -0,0 +1,139 @@
+name: sqa3d_sft_align_run1
+rng_seed: 42
+num_gpu: 4
+mode: train
+note: sqa3d_sft_align_run1
+naming_keywords:
+- dataloader.batchsize
+- task
+- note
+- time
+base_dir: results
+exp_dir: !!python/object/apply:pathlib.PosixPath
+- results
+- sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1
+- 2025-07-02-19:17:42.164767
+save_frequency: 10
+resume: false
+is_pretrain: false
+debug:
+  flag: false
+  hard_debug: false
+  debug_size: 20
+logger:
+  name: wandb
+  entity: yem
+  run_id: 5rulodim
+data:
+  train:
+  - ScanNetSQA3D
+  val:
+  - ScanNetSQA3D
+  test:
+  - ScanNetSQA3D
+  args:
+    max_obj_len: 80
+    max_seq_len: 50
+    num_points: 1024
+    pc_type: pred
+    sem_type: '607'
+    filter_lang: false
+    rot_aug: true
+  ScanNetSQA3D:
+    train:
+      use_unanswer: true
+    val:
+      use_unanswer: true
+    test:
+      use_unanswer: true
+  use_voxel: false
+  scan_family_base: existing_datasets/ScanNet
+  rscan_base: existing_datasets/3RScan
+task: SQA3D
+data_wrapper:
+  train: ScanFamilyDatasetWrapperOld
+  val: ScanFamilyDatasetWrapperOld
+  test: ScanFamilyDatasetWrapperOld
+trainer: DefaultTrainer
+ckpt_path: ''
+pretrain_ckpt_path: /home/m50048399/SceneVerse/results/sceneverse_scannet_exp1_b32_Pretrain_all_scannet_training_run1/b32_contrastive_mpm_random_caption/ckpt/ckpt_140.pth
+dataloader:
+  batchsize: 6
+  num_workers: 2
+  balance_dataset: false
+  filter_empty_annotations: false
+solver:
+  gradient_accumulation_steps: 1
+  epochs_per_save: 10
+  epochs_per_eval: 1
+  lr: 0.0001
+  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: SQA3DEval
+  save: false
+model:
+  name: OpenVocab
+  language:
+    name: BERTLanguageEncoder
+    args:
+      weights: bert-base-uncased
+      hidden_size: 768
+      num_hidden_layers: 4
+      num_attention_heads: 12
+      type_vocab_size: 2
+    lr: 1.0e-05
+  vision:
+    name: qihoo360/fg-clip-base
+    args:
+      backbone: pointnet++
+      hidden_size: 768
+      freeze: true
+      path: pretrained_weights/pointnetpp-open-bert
+      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: 0.0001
+  grounding:
+    name: UnifiedSpatialCrossEncoderV2
+    args:
+      hidden_size: 768
+      num_attention_heads: 12
+      num_layers: 4
+      dim_feedforward: 2048
+      dim_loc: 6
+    lr: 0.0001
+  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: 706
+  loss_type: ListLoss
+  loss_list:
+  - answer_loss
+  vis_loss_list:
+  - answer_loss

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:17:42.180036/config.yaml

@@ -0,0 +1,139 @@
+name: sqa3d_sft_align_run1
+rng_seed: 42
+num_gpu: 4
+mode: train
+note: sqa3d_sft_align_run1
+naming_keywords:
+- dataloader.batchsize
+- task
+- note
+- time
+base_dir: results
+exp_dir: !!python/object/apply:pathlib.PosixPath
+- results
+- sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1
+- 2025-07-02-19:17:42.180036
+save_frequency: 10
+resume: false
+is_pretrain: false
+debug:
+  flag: false
+  hard_debug: false
+  debug_size: 20
+logger:
+  name: wandb
+  entity: yem
+  run_id: e41lsz8r
+data:
+  train:
+  - ScanNetSQA3D
+  val:
+  - ScanNetSQA3D
+  test:
+  - ScanNetSQA3D
+  args:
+    max_obj_len: 80
+    max_seq_len: 50
+    num_points: 1024
+    pc_type: pred
+    sem_type: '607'
+    filter_lang: false
+    rot_aug: true
+  ScanNetSQA3D:
+    train:
+      use_unanswer: true
+    val:
+      use_unanswer: true
+    test:
+      use_unanswer: true
+  use_voxel: false
+  scan_family_base: existing_datasets/ScanNet
+  rscan_base: existing_datasets/3RScan
+task: SQA3D
+data_wrapper:
+  train: ScanFamilyDatasetWrapperOld
+  val: ScanFamilyDatasetWrapperOld
+  test: ScanFamilyDatasetWrapperOld
+trainer: DefaultTrainer
+ckpt_path: ''
+pretrain_ckpt_path: /home/m50048399/SceneVerse/results/sceneverse_scannet_exp1_b32_Pretrain_all_scannet_training_run1/b32_contrastive_mpm_random_caption/ckpt/ckpt_140.pth
+dataloader:
+  batchsize: 6
+  num_workers: 2
+  balance_dataset: false
+  filter_empty_annotations: false
+solver:
+  gradient_accumulation_steps: 1
+  epochs_per_save: 10
+  epochs_per_eval: 1
+  lr: 0.0001
+  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: SQA3DEval
+  save: false
+model:
+  name: OpenVocab
+  language:
+    name: BERTLanguageEncoder
+    args:
+      weights: bert-base-uncased
+      hidden_size: 768
+      num_hidden_layers: 4
+      num_attention_heads: 12
+      type_vocab_size: 2
+    lr: 1.0e-05
+  vision:
+    name: qihoo360/fg-clip-base
+    args:
+      backbone: pointnet++
+      hidden_size: 768
+      freeze: true
+      path: pretrained_weights/pointnetpp-open-bert
+      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: 0.0001
+  grounding:
+    name: UnifiedSpatialCrossEncoderV2
+    args:
+      hidden_size: 768
+      num_attention_heads: 12
+      num_layers: 4
+      dim_feedforward: 2048
+      dim_loc: 6
+    lr: 0.0001
+  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: 706
+  loss_type: ListLoss
+  loss_list:
+  - answer_loss
+  vis_loss_list:
+  - answer_loss

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:17:42.193410/config.yaml

@@ -0,0 +1,139 @@
+name: sqa3d_sft_align_run1
+rng_seed: 42
+num_gpu: 4
+mode: train
+note: sqa3d_sft_align_run1
+naming_keywords:
+- dataloader.batchsize
+- task
+- note
+- time
+base_dir: results
+exp_dir: !!python/object/apply:pathlib.PosixPath
+- results
+- sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1
+- 2025-07-02-19:17:42.193410
+save_frequency: 10
+resume: false
+is_pretrain: false
+debug:
+  flag: false
+  hard_debug: false
+  debug_size: 20
+logger:
+  name: wandb
+  entity: yem
+  run_id: aznum4ah
+data:
+  train:
+  - ScanNetSQA3D
+  val:
+  - ScanNetSQA3D
+  test:
+  - ScanNetSQA3D
+  args:
+    max_obj_len: 80
+    max_seq_len: 50
+    num_points: 1024
+    pc_type: pred
+    sem_type: '607'
+    filter_lang: false
+    rot_aug: true
+  ScanNetSQA3D:
+    train:
+      use_unanswer: true
+    val:
+      use_unanswer: true
+    test:
+      use_unanswer: true
+  use_voxel: false
+  scan_family_base: existing_datasets/ScanNet
+  rscan_base: existing_datasets/3RScan
+task: SQA3D
+data_wrapper:
+  train: ScanFamilyDatasetWrapperOld
+  val: ScanFamilyDatasetWrapperOld
+  test: ScanFamilyDatasetWrapperOld
+trainer: DefaultTrainer
+ckpt_path: ''
+pretrain_ckpt_path: /home/m50048399/SceneVerse/results/sceneverse_scannet_exp1_b32_Pretrain_all_scannet_training_run1/b32_contrastive_mpm_random_caption/ckpt/ckpt_140.pth
+dataloader:
+  batchsize: 6
+  num_workers: 2
+  balance_dataset: false
+  filter_empty_annotations: false
+solver:
+  gradient_accumulation_steps: 1
+  epochs_per_save: 10
+  epochs_per_eval: 1
+  lr: 0.0001
+  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: SQA3DEval
+  save: false
+model:
+  name: OpenVocab
+  language:
+    name: BERTLanguageEncoder
+    args:
+      weights: bert-base-uncased
+      hidden_size: 768
+      num_hidden_layers: 4
+      num_attention_heads: 12
+      type_vocab_size: 2
+    lr: 1.0e-05
+  vision:
+    name: qihoo360/fg-clip-base
+    args:
+      backbone: pointnet++
+      hidden_size: 768
+      freeze: true
+      path: pretrained_weights/pointnetpp-open-bert
+      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: 0.0001
+  grounding:
+    name: UnifiedSpatialCrossEncoderV2
+    args:
+      hidden_size: 768
+      num_attention_heads: 12
+      num_layers: 4
+      dim_feedforward: 2048
+      dim_loc: 6
+    lr: 0.0001
+  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: 706
+  loss_type: ListLoss
+  loss_list:
+  - answer_loss
+  vis_loss_list:
+  - answer_loss

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:40:37.831550/config.yaml

@@ -0,0 +1,139 @@
+name: sqa3d_sft_align_run1
+rng_seed: 42
+num_gpu: 4
+mode: train
+note: sqa3d_sft_align_run1
+naming_keywords:
+- dataloader.batchsize
+- task
+- note
+- time
+base_dir: results
+exp_dir: !!python/object/apply:pathlib.PosixPath
+- results
+- sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1
+- 2025-07-02-19:40:37.831550
+save_frequency: 10
+resume: false
+is_pretrain: false
+debug:
+  flag: false
+  hard_debug: false
+  debug_size: 20
+logger:
+  name: wandb
+  entity: yem
+  run_id: mzsgpyqt
+data:
+  train:
+  - ScanNetSQA3D
+  val:
+  - ScanNetSQA3D
+  test:
+  - ScanNetSQA3D
+  args:
+    max_obj_len: 80
+    max_seq_len: 50
+    num_points: 1024
+    pc_type: pred
+    sem_type: '607'
+    filter_lang: false
+    rot_aug: true
+  ScanNetSQA3D:
+    train:
+      use_unanswer: true
+    val:
+      use_unanswer: true
+    test:
+      use_unanswer: true
+  use_voxel: false
+  scan_family_base: existing_datasets/ScanNet
+  rscan_base: existing_datasets/3RScan
+task: SQA3D
+data_wrapper:
+  train: ScanFamilyDatasetWrapperOld
+  val: ScanFamilyDatasetWrapperOld
+  test: ScanFamilyDatasetWrapperOld
+trainer: DefaultTrainer
+ckpt_path: ''
+pretrain_ckpt_path: /home/m50048399/SceneVerse/results/sceneverse_scannet_exp1_b32_Pretrain_all_scannet_training_run1/b32_contrastive_mpm_random_caption/ckpt/ckpt_140.pth
+dataloader:
+  batchsize: 6
+  num_workers: 2
+  balance_dataset: false
+  filter_empty_annotations: false
+solver:
+  gradient_accumulation_steps: 1
+  epochs_per_save: 10
+  epochs_per_eval: 1
+  lr: 0.0001
+  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: SQA3DEval
+  save: false
+model:
+  name: OpenVocab
+  language:
+    name: BERTLanguageEncoder
+    args:
+      weights: bert-base-uncased
+      hidden_size: 768
+      num_hidden_layers: 4
+      num_attention_heads: 12
+      type_vocab_size: 2
+    lr: 1.0e-05
+  vision:
+    name: qihoo360/fg-clip-base
+    args:
+      backbone: pointnet++
+      hidden_size: 768
+      freeze: true
+      path: pretrained_weights/pointnetpp-open-bert
+      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: 0.0001
+  grounding:
+    name: UnifiedSpatialCrossEncoderV2
+    args:
+      hidden_size: 768
+      num_attention_heads: 12
+      num_layers: 4
+      dim_feedforward: 2048
+      dim_loc: 6
+    lr: 0.0001
+  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: 706
+  loss_type: ListLoss
+  loss_list:
+  - answer_loss
+  vis_loss_list:
+  - answer_loss

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:40:37.833170/ckpt/best.pth/custom_checkpoint_0.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eba7c6f2d33d65a9f1284fa6cab462224dbaacc4469d2c4332b498ecf2fd14ca
+size 1040

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:40:37.833170/ckpt/best.pth/model_1.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:60d95b10b6e140a9626a7058d5038528f2ff80148dc4569b881db56052046509
+size 40

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:40:37.833170/ckpt/best.pth/random_states_0.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e2d29f92a09daef7a7c2cc2fad629d0c329ba7c3bc3cd1c94a9bd32d0782d52
+size 15124

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:40:37.833170/ckpt/best.pth/scheduler.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a02ba49a043e4028cddd88477a825c75b63a511de04c5483cd6fe1191fbc986
+size 1064

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:40:37.833170/ckpt/latest.pth/custom_checkpoint_0.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eba7c6f2d33d65a9f1284fa6cab462224dbaacc4469d2c4332b498ecf2fd14ca
+size 1040

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:40:37.833170/ckpt/latest.pth/model_1.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:60d95b10b6e140a9626a7058d5038528f2ff80148dc4569b881db56052046509
+size 40

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:40:37.833170/ckpt/latest.pth/random_states_0.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e2d29f92a09daef7a7c2cc2fad629d0c329ba7c3bc3cd1c94a9bd32d0782d52
+size 15124

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:40:37.833170/ckpt/latest.pth/scheduler.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a02ba49a043e4028cddd88477a825c75b63a511de04c5483cd6fe1191fbc986
+size 1064

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:40:37.833170/config.yaml

@@ -0,0 +1,139 @@
+name: sqa3d_sft_align_run1
+rng_seed: 42
+num_gpu: 4
+mode: train
+note: sqa3d_sft_align_run1
+naming_keywords:
+- dataloader.batchsize
+- task
+- note
+- time
+base_dir: results
+exp_dir: !!python/object/apply:pathlib.PosixPath
+- results
+- sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1
+- 2025-07-02-19:40:37.833170
+save_frequency: 10
+resume: false
+is_pretrain: false
+debug:
+  flag: false
+  hard_debug: false
+  debug_size: 20
+logger:
+  name: wandb
+  entity: yem
+  run_id: 9ytj0xsm
+data:
+  train:
+  - ScanNetSQA3D
+  val:
+  - ScanNetSQA3D
+  test:
+  - ScanNetSQA3D
+  args:
+    max_obj_len: 80
+    max_seq_len: 50
+    num_points: 1024
+    pc_type: pred
+    sem_type: '607'
+    filter_lang: false
+    rot_aug: true
+  ScanNetSQA3D:
+    train:
+      use_unanswer: true
+    val:
+      use_unanswer: true
+    test:
+      use_unanswer: true
+  use_voxel: false
+  scan_family_base: existing_datasets/ScanNet
+  rscan_base: existing_datasets/3RScan
+task: SQA3D
+data_wrapper:
+  train: ScanFamilyDatasetWrapperOld
+  val: ScanFamilyDatasetWrapperOld
+  test: ScanFamilyDatasetWrapperOld
+trainer: DefaultTrainer
+ckpt_path: ''
+pretrain_ckpt_path: /home/m50048399/SceneVerse/results/sceneverse_scannet_exp1_b32_Pretrain_all_scannet_training_run1/b32_contrastive_mpm_random_caption/ckpt/ckpt_140.pth
+dataloader:
+  batchsize: 6
+  num_workers: 2
+  balance_dataset: false
+  filter_empty_annotations: false
+solver:
+  gradient_accumulation_steps: 1
+  epochs_per_save: 10
+  epochs_per_eval: 1
+  lr: 0.0001
+  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: SQA3DEval
+  save: false
+model:
+  name: OpenVocab
+  language:
+    name: BERTLanguageEncoder
+    args:
+      weights: bert-base-uncased
+      hidden_size: 768
+      num_hidden_layers: 4
+      num_attention_heads: 12
+      type_vocab_size: 2
+    lr: 1.0e-05
+  vision:
+    name: qihoo360/fg-clip-base
+    args:
+      backbone: pointnet++
+      hidden_size: 768
+      freeze: true
+      path: pretrained_weights/pointnetpp-open-bert
+      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: 0.0001
+  grounding:
+    name: UnifiedSpatialCrossEncoderV2
+    args:
+      hidden_size: 768
+      num_attention_heads: 12
+      num_layers: 4
+      dim_feedforward: 2048
+      dim_loc: 6
+    lr: 0.0001
+  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: 706
+  loss_type: ListLoss
+  loss_list:
+  - answer_loss
+  vis_loss_list:
+  - answer_loss

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:40:37.834617/config.yaml

@@ -0,0 +1,139 @@
+name: sqa3d_sft_align_run1
+rng_seed: 42
+num_gpu: 4
+mode: train
+note: sqa3d_sft_align_run1
+naming_keywords:
+- dataloader.batchsize
+- task
+- note
+- time
+base_dir: results
+exp_dir: !!python/object/apply:pathlib.PosixPath
+- results
+- sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1
+- 2025-07-02-19:40:37.834617
+save_frequency: 10
+resume: false
+is_pretrain: false
+debug:
+  flag: false
+  hard_debug: false
+  debug_size: 20
+logger:
+  name: wandb
+  entity: yem
+  run_id: 8mgu97vy
+data:
+  train:
+  - ScanNetSQA3D
+  val:
+  - ScanNetSQA3D
+  test:
+  - ScanNetSQA3D
+  args:
+    max_obj_len: 80
+    max_seq_len: 50
+    num_points: 1024
+    pc_type: pred
+    sem_type: '607'
+    filter_lang: false
+    rot_aug: true
+  ScanNetSQA3D:
+    train:
+      use_unanswer: true
+    val:
+      use_unanswer: true
+    test:
+      use_unanswer: true
+  use_voxel: false
+  scan_family_base: existing_datasets/ScanNet
+  rscan_base: existing_datasets/3RScan
+task: SQA3D
+data_wrapper:
+  train: ScanFamilyDatasetWrapperOld
+  val: ScanFamilyDatasetWrapperOld
+  test: ScanFamilyDatasetWrapperOld
+trainer: DefaultTrainer
+ckpt_path: ''
+pretrain_ckpt_path: /home/m50048399/SceneVerse/results/sceneverse_scannet_exp1_b32_Pretrain_all_scannet_training_run1/b32_contrastive_mpm_random_caption/ckpt/ckpt_140.pth
+dataloader:
+  batchsize: 6
+  num_workers: 2
+  balance_dataset: false
+  filter_empty_annotations: false
+solver:
+  gradient_accumulation_steps: 1
+  epochs_per_save: 10
+  epochs_per_eval: 1
+  lr: 0.0001
+  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: SQA3DEval
+  save: false
+model:
+  name: OpenVocab
+  language:
+    name: BERTLanguageEncoder
+    args:
+      weights: bert-base-uncased
+      hidden_size: 768
+      num_hidden_layers: 4
+      num_attention_heads: 12
+      type_vocab_size: 2
+    lr: 1.0e-05
+  vision:
+    name: qihoo360/fg-clip-base
+    args:
+      backbone: pointnet++
+      hidden_size: 768
+      freeze: true
+      path: pretrained_weights/pointnetpp-open-bert
+      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: 0.0001
+  grounding:
+    name: UnifiedSpatialCrossEncoderV2
+    args:
+      hidden_size: 768
+      num_attention_heads: 12
+      num_layers: 4
+      dim_feedforward: 2048
+      dim_loc: 6
+    lr: 0.0001
+  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: 706
+  loss_type: ListLoss
+  loss_list:
+  - answer_loss
+  vis_loss_list:
+  - answer_loss

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:40:37.836494/config.yaml

@@ -0,0 +1,139 @@
+name: sqa3d_sft_align_run1
+rng_seed: 42
+num_gpu: 4
+mode: train
+note: sqa3d_sft_align_run1
+naming_keywords:
+- dataloader.batchsize
+- task
+- note
+- time
+base_dir: results
+exp_dir: !!python/object/apply:pathlib.PosixPath
+- results
+- sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1
+- 2025-07-02-19:40:37.836494
+save_frequency: 10
+resume: false
+is_pretrain: false
+debug:
+  flag: false
+  hard_debug: false
+  debug_size: 20
+logger:
+  name: wandb
+  entity: yem
+  run_id: uew009q3
+data:
+  train:
+  - ScanNetSQA3D
+  val:
+  - ScanNetSQA3D
+  test:
+  - ScanNetSQA3D
+  args:
+    max_obj_len: 80
+    max_seq_len: 50
+    num_points: 1024
+    pc_type: pred
+    sem_type: '607'
+    filter_lang: false
+    rot_aug: true
+  ScanNetSQA3D:
+    train:
+      use_unanswer: true
+    val:
+      use_unanswer: true
+    test:
+      use_unanswer: true
+  use_voxel: false
+  scan_family_base: existing_datasets/ScanNet
+  rscan_base: existing_datasets/3RScan
+task: SQA3D
+data_wrapper:
+  train: ScanFamilyDatasetWrapperOld
+  val: ScanFamilyDatasetWrapperOld
+  test: ScanFamilyDatasetWrapperOld
+trainer: DefaultTrainer
+ckpt_path: ''
+pretrain_ckpt_path: /home/m50048399/SceneVerse/results/sceneverse_scannet_exp1_b32_Pretrain_all_scannet_training_run1/b32_contrastive_mpm_random_caption/ckpt/ckpt_140.pth
+dataloader:
+  batchsize: 6
+  num_workers: 2
+  balance_dataset: false
+  filter_empty_annotations: false
+solver:
+  gradient_accumulation_steps: 1
+  epochs_per_save: 10
+  epochs_per_eval: 1
+  lr: 0.0001
+  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: SQA3DEval
+  save: false
+model:
+  name: OpenVocab
+  language:
+    name: BERTLanguageEncoder
+    args:
+      weights: bert-base-uncased
+      hidden_size: 768
+      num_hidden_layers: 4
+      num_attention_heads: 12
+      type_vocab_size: 2
+    lr: 1.0e-05
+  vision:
+    name: qihoo360/fg-clip-base
+    args:
+      backbone: pointnet++
+      hidden_size: 768
+      freeze: true
+      path: pretrained_weights/pointnetpp-open-bert
+      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: 0.0001
+  grounding:
+    name: UnifiedSpatialCrossEncoderV2
+    args:
+      hidden_size: 768
+      num_attention_heads: 12
+      num_layers: 4
+      dim_feedforward: 2048
+      dim_loc: 6
+    lr: 0.0001
+  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: 706
+  loss_type: ListLoss
+  loss_list:
+  - answer_loss
+  vis_loss_list:
+  - answer_loss

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:55:49.009678/config.yaml

@@ -0,0 +1,139 @@
+name: sqa3d_sft_align_run1
+rng_seed: 42
+num_gpu: 4
+mode: train
+note: sqa3d_sft_align_run1
+naming_keywords:
+- dataloader.batchsize
+- task
+- note
+- time
+base_dir: results
+exp_dir: !!python/object/apply:pathlib.PosixPath
+- results
+- sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1
+- 2025-07-02-19:55:49.009678
+save_frequency: 10
+resume: false
+is_pretrain: false
+debug:
+  flag: false
+  hard_debug: false
+  debug_size: 20
+logger:
+  name: wandb
+  entity: yem
+  run_id: 6vt24e5s
+data:
+  train:
+  - ScanNetSQA3D
+  val:
+  - ScanNetSQA3D
+  test:
+  - ScanNetSQA3D
+  args:
+    max_obj_len: 80
+    max_seq_len: 50
+    num_points: 1024
+    pc_type: pred
+    sem_type: '607'
+    filter_lang: false
+    rot_aug: true
+  ScanNetSQA3D:
+    train:
+      use_unanswer: true
+    val:
+      use_unanswer: true
+    test:
+      use_unanswer: true
+  use_voxel: false
+  scan_family_base: existing_datasets/ScanNet
+  rscan_base: existing_datasets/3RScan
+task: SQA3D
+data_wrapper:
+  train: ScanFamilyDatasetWrapperOld
+  val: ScanFamilyDatasetWrapperOld
+  test: ScanFamilyDatasetWrapperOld
+trainer: DefaultTrainer
+ckpt_path: ''
+pretrain_ckpt_path: ''
+dataloader:
+  batchsize: 6
+  num_workers: 2
+  balance_dataset: false
+  filter_empty_annotations: false
+solver:
+  gradient_accumulation_steps: 1
+  epochs_per_save: 10
+  epochs_per_eval: 1
+  lr: 0.0001
+  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: SQA3DEval
+  save: false
+model:
+  name: OpenVocab
+  language:
+    name: BERTLanguageEncoder
+    args:
+      weights: bert-base-uncased
+      hidden_size: 768
+      num_hidden_layers: 4
+      num_attention_heads: 12
+      type_vocab_size: 2
+    lr: 1.0e-05
+  vision:
+    name: qihoo360/fg-clip-base
+    args:
+      backbone: pointnet++
+      hidden_size: 768
+      freeze: true
+      path: pretrained_weights/pointnetpp-open-bert
+      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: 0.0001
+  grounding:
+    name: UnifiedSpatialCrossEncoderV2
+    args:
+      hidden_size: 768
+      num_attention_heads: 12
+      num_layers: 4
+      dim_feedforward: 2048
+      dim_loc: 6
+    lr: 0.0001
+  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: 706
+  loss_type: ListLoss
+  loss_list:
+  - answer_loss
+  vis_loss_list:
+  - answer_loss

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:55:49.014082/config.yaml

@@ -0,0 +1,139 @@
+name: sqa3d_sft_align_run1
+rng_seed: 42
+num_gpu: 4
+mode: train
+note: sqa3d_sft_align_run1
+naming_keywords:
+- dataloader.batchsize
+- task
+- note
+- time
+base_dir: results
+exp_dir: !!python/object/apply:pathlib.PosixPath
+- results
+- sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1
+- 2025-07-02-19:55:49.014082
+save_frequency: 10
+resume: false
+is_pretrain: false
+debug:
+  flag: false
+  hard_debug: false
+  debug_size: 20
+logger:
+  name: wandb
+  entity: yem
+  run_id: 3t3mseu2
+data:
+  train:
+  - ScanNetSQA3D
+  val:
+  - ScanNetSQA3D
+  test:
+  - ScanNetSQA3D
+  args:
+    max_obj_len: 80
+    max_seq_len: 50
+    num_points: 1024
+    pc_type: pred
+    sem_type: '607'
+    filter_lang: false
+    rot_aug: true
+  ScanNetSQA3D:
+    train:
+      use_unanswer: true
+    val:
+      use_unanswer: true
+    test:
+      use_unanswer: true
+  use_voxel: false
+  scan_family_base: existing_datasets/ScanNet
+  rscan_base: existing_datasets/3RScan
+task: SQA3D
+data_wrapper:
+  train: ScanFamilyDatasetWrapperOld
+  val: ScanFamilyDatasetWrapperOld
+  test: ScanFamilyDatasetWrapperOld
+trainer: DefaultTrainer
+ckpt_path: ''
+pretrain_ckpt_path: ''
+dataloader:
+  batchsize: 6
+  num_workers: 2
+  balance_dataset: false
+  filter_empty_annotations: false
+solver:
+  gradient_accumulation_steps: 1
+  epochs_per_save: 10
+  epochs_per_eval: 1
+  lr: 0.0001
+  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: SQA3DEval
+  save: false
+model:
+  name: OpenVocab
+  language:
+    name: BERTLanguageEncoder
+    args:
+      weights: bert-base-uncased
+      hidden_size: 768
+      num_hidden_layers: 4
+      num_attention_heads: 12
+      type_vocab_size: 2
+    lr: 1.0e-05
+  vision:
+    name: qihoo360/fg-clip-base
+    args:
+      backbone: pointnet++
+      hidden_size: 768
+      freeze: true
+      path: pretrained_weights/pointnetpp-open-bert
+      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: 0.0001
+  grounding:
+    name: UnifiedSpatialCrossEncoderV2
+    args:
+      hidden_size: 768
+      num_attention_heads: 12
+      num_layers: 4
+      dim_feedforward: 2048
+      dim_loc: 6
+    lr: 0.0001
+  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: 706
+  loss_type: ListLoss
+  loss_list:
+  - answer_loss
+  vis_loss_list:
+  - answer_loss

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:55:49.043158/config.yaml

@@ -0,0 +1,139 @@
+name: sqa3d_sft_align_run1
+rng_seed: 42
+num_gpu: 4
+mode: train
+note: sqa3d_sft_align_run1
+naming_keywords:
+- dataloader.batchsize
+- task
+- note
+- time
+base_dir: results
+exp_dir: !!python/object/apply:pathlib.PosixPath
+- results
+- sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1
+- 2025-07-02-19:55:49.043158
+save_frequency: 10
+resume: false
+is_pretrain: false
+debug:
+  flag: false
+  hard_debug: false
+  debug_size: 20
+logger:
+  name: wandb
+  entity: yem
+  run_id: 6ey0yefb
+data:
+  train:
+  - ScanNetSQA3D
+  val:
+  - ScanNetSQA3D
+  test:
+  - ScanNetSQA3D
+  args:
+    max_obj_len: 80
+    max_seq_len: 50
+    num_points: 1024
+    pc_type: pred
+    sem_type: '607'
+    filter_lang: false
+    rot_aug: true
+  ScanNetSQA3D:
+    train:
+      use_unanswer: true
+    val:
+      use_unanswer: true
+    test:
+      use_unanswer: true
+  use_voxel: false
+  scan_family_base: existing_datasets/ScanNet
+  rscan_base: existing_datasets/3RScan
+task: SQA3D
+data_wrapper:
+  train: ScanFamilyDatasetWrapperOld
+  val: ScanFamilyDatasetWrapperOld
+  test: ScanFamilyDatasetWrapperOld
+trainer: DefaultTrainer
+ckpt_path: ''
+pretrain_ckpt_path: ''
+dataloader:
+  batchsize: 6
+  num_workers: 2
+  balance_dataset: false
+  filter_empty_annotations: false
+solver:
+  gradient_accumulation_steps: 1
+  epochs_per_save: 10
+  epochs_per_eval: 1
+  lr: 0.0001
+  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: SQA3DEval
+  save: false
+model:
+  name: OpenVocab
+  language:
+    name: BERTLanguageEncoder
+    args:
+      weights: bert-base-uncased
+      hidden_size: 768
+      num_hidden_layers: 4
+      num_attention_heads: 12
+      type_vocab_size: 2
+    lr: 1.0e-05
+  vision:
+    name: qihoo360/fg-clip-base
+    args:
+      backbone: pointnet++
+      hidden_size: 768
+      freeze: true
+      path: pretrained_weights/pointnetpp-open-bert
+      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: 0.0001
+  grounding:
+    name: UnifiedSpatialCrossEncoderV2
+    args:
+      hidden_size: 768
+      num_attention_heads: 12
+      num_layers: 4
+      dim_feedforward: 2048
+      dim_loc: 6
+    lr: 0.0001
+  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: 706
+  loss_type: ListLoss
+  loss_list:
+  - answer_loss
+  vis_loss_list:
+  - answer_loss

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:55:49.046517/config.yaml

@@ -0,0 +1,139 @@
+name: sqa3d_sft_align_run1
+rng_seed: 42
+num_gpu: 4
+mode: train
+note: sqa3d_sft_align_run1
+naming_keywords:
+- dataloader.batchsize
+- task
+- note
+- time
+base_dir: results
+exp_dir: !!python/object/apply:pathlib.PosixPath
+- results
+- sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1
+- 2025-07-02-19:55:49.046517
+save_frequency: 10
+resume: false
+is_pretrain: false
+debug:
+  flag: false
+  hard_debug: false
+  debug_size: 20
+logger:
+  name: wandb
+  entity: yem
+  run_id: 6abslpit
+data:
+  train:
+  - ScanNetSQA3D
+  val:
+  - ScanNetSQA3D
+  test:
+  - ScanNetSQA3D
+  args:
+    max_obj_len: 80
+    max_seq_len: 50
+    num_points: 1024
+    pc_type: pred
+    sem_type: '607'
+    filter_lang: false
+    rot_aug: true
+  ScanNetSQA3D:
+    train:
+      use_unanswer: true
+    val:
+      use_unanswer: true
+    test:
+      use_unanswer: true
+  use_voxel: false
+  scan_family_base: existing_datasets/ScanNet
+  rscan_base: existing_datasets/3RScan
+task: SQA3D
+data_wrapper:
+  train: ScanFamilyDatasetWrapperOld
+  val: ScanFamilyDatasetWrapperOld
+  test: ScanFamilyDatasetWrapperOld
+trainer: DefaultTrainer
+ckpt_path: ''
+pretrain_ckpt_path: ''
+dataloader:
+  batchsize: 6
+  num_workers: 2
+  balance_dataset: false
+  filter_empty_annotations: false
+solver:
+  gradient_accumulation_steps: 1
+  epochs_per_save: 10
+  epochs_per_eval: 1
+  lr: 0.0001
+  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: SQA3DEval
+  save: false
+model:
+  name: OpenVocab
+  language:
+    name: BERTLanguageEncoder
+    args:
+      weights: bert-base-uncased
+      hidden_size: 768
+      num_hidden_layers: 4
+      num_attention_heads: 12
+      type_vocab_size: 2
+    lr: 1.0e-05
+  vision:
+    name: qihoo360/fg-clip-base
+    args:
+      backbone: pointnet++
+      hidden_size: 768
+      freeze: true
+      path: pretrained_weights/pointnetpp-open-bert
+      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: 0.0001
+  grounding:
+    name: UnifiedSpatialCrossEncoderV2
+    args:
+      hidden_size: 768
+      num_attention_heads: 12
+      num_layers: 4
+      dim_feedforward: 2048
+      dim_loc: 6
+    lr: 0.0001
+  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: 706
+  loss_type: ListLoss
+  loss_list:
+  - answer_loss
+  vis_loss_list:
+  - answer_loss

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:59:21.609689/config.yaml

@@ -0,0 +1,139 @@
+name: sqa3d_sft_align_run1
+rng_seed: 42
+num_gpu: 4
+mode: train
+note: sqa3d_sft_align_run1
+naming_keywords:
+- dataloader.batchsize
+- task
+- note
+- time
+base_dir: results
+exp_dir: !!python/object/apply:pathlib.PosixPath
+- results
+- sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1
+- 2025-07-02-19:59:21.609689
+save_frequency: 10
+resume: false
+is_pretrain: false
+debug:
+  flag: false
+  hard_debug: false
+  debug_size: 20
+logger:
+  name: wandb
+  entity: yem
+  run_id: quf22pv9
+data:
+  train:
+  - ScanNetSQA3D
+  val:
+  - ScanNetSQA3D
+  test:
+  - ScanNetSQA3D
+  args:
+    max_obj_len: 80
+    max_seq_len: 50
+    num_points: 1024
+    pc_type: pred
+    sem_type: '607'
+    filter_lang: false
+    rot_aug: true
+  ScanNetSQA3D:
+    train:
+      use_unanswer: true
+    val:
+      use_unanswer: true
+    test:
+      use_unanswer: true
+  use_voxel: false
+  scan_family_base: existing_datasets/ScanNet
+  rscan_base: existing_datasets/3RScan
+task: SQA3D
+data_wrapper:
+  train: ScanFamilyDatasetWrapperOld
+  val: ScanFamilyDatasetWrapperOld
+  test: ScanFamilyDatasetWrapperOld
+trainer: DefaultTrainer
+ckpt_path: ''
+pretrain_ckpt_path: ''
+dataloader:
+  batchsize: 6
+  num_workers: 2
+  balance_dataset: false
+  filter_empty_annotations: false
+solver:
+  gradient_accumulation_steps: 1
+  epochs_per_save: 10
+  epochs_per_eval: 1
+  lr: 0.0001
+  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: SQA3DEval
+  save: false
+model:
+  name: OpenVocab
+  language:
+    name: BERTLanguageEncoder
+    args:
+      weights: bert-base-uncased
+      hidden_size: 768
+      num_hidden_layers: 4
+      num_attention_heads: 12
+      type_vocab_size: 2
+    lr: 1.0e-05
+  vision:
+    name: qihoo360/fg-clip-base
+    args:
+      backbone: pointnet++
+      hidden_size: 768
+      freeze: true
+      path: pretrained_weights/pointnetpp-open-bert
+      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: 0.0001
+  grounding:
+    name: UnifiedSpatialCrossEncoderV2
+    args:
+      hidden_size: 768
+      num_attention_heads: 12
+      num_layers: 4
+      dim_feedforward: 2048
+      dim_loc: 6
+    lr: 0.0001
+  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: 706
+  loss_type: ListLoss
+  loss_list:
+  - answer_loss
+  vis_loss_list:
+  - answer_loss

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:59:21.612584/config.yaml

@@ -0,0 +1,139 @@
+name: sqa3d_sft_align_run1
+rng_seed: 42
+num_gpu: 4
+mode: train
+note: sqa3d_sft_align_run1
+naming_keywords:
+- dataloader.batchsize
+- task
+- note
+- time
+base_dir: results
+exp_dir: !!python/object/apply:pathlib.PosixPath
+- results
+- sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1
+- 2025-07-02-19:59:21.612584
+save_frequency: 10
+resume: false
+is_pretrain: false
+debug:
+  flag: false
+  hard_debug: false
+  debug_size: 20
+logger:
+  name: wandb
+  entity: yem
+  run_id: jb2ua7xw
+data:
+  train:
+  - ScanNetSQA3D
+  val:
+  - ScanNetSQA3D
+  test:
+  - ScanNetSQA3D
+  args:
+    max_obj_len: 80
+    max_seq_len: 50
+    num_points: 1024
+    pc_type: pred
+    sem_type: '607'
+    filter_lang: false
+    rot_aug: true
+  ScanNetSQA3D:
+    train:
+      use_unanswer: true
+    val:
+      use_unanswer: true
+    test:
+      use_unanswer: true
+  use_voxel: false
+  scan_family_base: existing_datasets/ScanNet
+  rscan_base: existing_datasets/3RScan
+task: SQA3D
+data_wrapper:
+  train: ScanFamilyDatasetWrapperOld
+  val: ScanFamilyDatasetWrapperOld
+  test: ScanFamilyDatasetWrapperOld
+trainer: DefaultTrainer
+ckpt_path: ''
+pretrain_ckpt_path: ''
+dataloader:
+  batchsize: 6
+  num_workers: 2
+  balance_dataset: false
+  filter_empty_annotations: false
+solver:
+  gradient_accumulation_steps: 1
+  epochs_per_save: 10
+  epochs_per_eval: 1
+  lr: 0.0001
+  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: SQA3DEval
+  save: false
+model:
+  name: OpenVocab
+  language:
+    name: BERTLanguageEncoder
+    args:
+      weights: bert-base-uncased
+      hidden_size: 768
+      num_hidden_layers: 4
+      num_attention_heads: 12
+      type_vocab_size: 2
+    lr: 1.0e-05
+  vision:
+    name: qihoo360/fg-clip-base
+    args:
+      backbone: pointnet++
+      hidden_size: 768
+      freeze: true
+      path: pretrained_weights/pointnetpp-open-bert
+      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: 0.0001
+  grounding:
+    name: UnifiedSpatialCrossEncoderV2
+    args:
+      hidden_size: 768
+      num_attention_heads: 12
+      num_layers: 4
+      dim_feedforward: 2048
+      dim_loc: 6
+    lr: 0.0001
+  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: 706
+  loss_type: ListLoss
+  loss_list:
+  - answer_loss
+  vis_loss_list:
+  - answer_loss

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:59:21.618178/config.yaml

@@ -0,0 +1,139 @@
+name: sqa3d_sft_align_run1
+rng_seed: 42
+num_gpu: 4
+mode: train
+note: sqa3d_sft_align_run1
+naming_keywords:
+- dataloader.batchsize
+- task
+- note
+- time
+base_dir: results
+exp_dir: !!python/object/apply:pathlib.PosixPath
+- results
+- sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1
+- 2025-07-02-19:59:21.618178
+save_frequency: 10
+resume: false
+is_pretrain: false
+debug:
+  flag: false
+  hard_debug: false
+  debug_size: 20
+logger:
+  name: wandb
+  entity: yem
+  run_id: as519d0q
+data:
+  train:
+  - ScanNetSQA3D
+  val:
+  - ScanNetSQA3D
+  test:
+  - ScanNetSQA3D
+  args:
+    max_obj_len: 80
+    max_seq_len: 50
+    num_points: 1024
+    pc_type: pred
+    sem_type: '607'
+    filter_lang: false
+    rot_aug: true
+  ScanNetSQA3D:
+    train:
+      use_unanswer: true
+    val:
+      use_unanswer: true
+    test:
+      use_unanswer: true
+  use_voxel: false
+  scan_family_base: existing_datasets/ScanNet
+  rscan_base: existing_datasets/3RScan
+task: SQA3D
+data_wrapper:
+  train: ScanFamilyDatasetWrapperOld
+  val: ScanFamilyDatasetWrapperOld
+  test: ScanFamilyDatasetWrapperOld
+trainer: DefaultTrainer
+ckpt_path: ''
+pretrain_ckpt_path: ''
+dataloader:
+  batchsize: 6
+  num_workers: 2
+  balance_dataset: false
+  filter_empty_annotations: false
+solver:
+  gradient_accumulation_steps: 1
+  epochs_per_save: 10
+  epochs_per_eval: 1
+  lr: 0.0001
+  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: SQA3DEval
+  save: false
+model:
+  name: OpenVocab
+  language:
+    name: BERTLanguageEncoder
+    args:
+      weights: bert-base-uncased
+      hidden_size: 768
+      num_hidden_layers: 4
+      num_attention_heads: 12
+      type_vocab_size: 2
+    lr: 1.0e-05
+  vision:
+    name: qihoo360/fg-clip-base
+    args:
+      backbone: pointnet++
+      hidden_size: 768
+      freeze: true
+      path: pretrained_weights/pointnetpp-open-bert
+      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: 0.0001
+  grounding:
+    name: UnifiedSpatialCrossEncoderV2
+    args:
+      hidden_size: 768
+      num_attention_heads: 12
+      num_layers: 4
+      dim_feedforward: 2048
+      dim_loc: 6
+    lr: 0.0001
+  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: 706
+  loss_type: ListLoss
+  loss_list:
+  - answer_loss
+  vis_loss_list:
+  - answer_loss

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-19:59:21.638340/config.yaml

@@ -0,0 +1,139 @@
+name: sqa3d_sft_align_run1
+rng_seed: 42
+num_gpu: 4
+mode: train
+note: sqa3d_sft_align_run1
+naming_keywords:
+- dataloader.batchsize
+- task
+- note
+- time
+base_dir: results
+exp_dir: !!python/object/apply:pathlib.PosixPath
+- results
+- sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1
+- 2025-07-02-19:59:21.638340
+save_frequency: 10
+resume: false
+is_pretrain: false
+debug:
+  flag: false
+  hard_debug: false
+  debug_size: 20
+logger:
+  name: wandb
+  entity: yem
+  run_id: bhv4qvbz
+data:
+  train:
+  - ScanNetSQA3D
+  val:
+  - ScanNetSQA3D
+  test:
+  - ScanNetSQA3D
+  args:
+    max_obj_len: 80
+    max_seq_len: 50
+    num_points: 1024
+    pc_type: pred
+    sem_type: '607'
+    filter_lang: false
+    rot_aug: true
+  ScanNetSQA3D:
+    train:
+      use_unanswer: true
+    val:
+      use_unanswer: true
+    test:
+      use_unanswer: true
+  use_voxel: false
+  scan_family_base: existing_datasets/ScanNet
+  rscan_base: existing_datasets/3RScan
+task: SQA3D
+data_wrapper:
+  train: ScanFamilyDatasetWrapperOld
+  val: ScanFamilyDatasetWrapperOld
+  test: ScanFamilyDatasetWrapperOld
+trainer: DefaultTrainer
+ckpt_path: ''
+pretrain_ckpt_path: ''
+dataloader:
+  batchsize: 6
+  num_workers: 2
+  balance_dataset: false
+  filter_empty_annotations: false
+solver:
+  gradient_accumulation_steps: 1
+  epochs_per_save: 10
+  epochs_per_eval: 1
+  lr: 0.0001
+  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: SQA3DEval
+  save: false
+model:
+  name: OpenVocab
+  language:
+    name: BERTLanguageEncoder
+    args:
+      weights: bert-base-uncased
+      hidden_size: 768
+      num_hidden_layers: 4
+      num_attention_heads: 12
+      type_vocab_size: 2
+    lr: 1.0e-05
+  vision:
+    name: qihoo360/fg-clip-base
+    args:
+      backbone: pointnet++
+      hidden_size: 768
+      freeze: true
+      path: pretrained_weights/pointnetpp-open-bert
+      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: 0.0001
+  grounding:
+    name: UnifiedSpatialCrossEncoderV2
+    args:
+      hidden_size: 768
+      num_attention_heads: 12
+      num_layers: 4
+      dim_feedforward: 2048
+      dim_loc: 6
+    lr: 0.0001
+  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: 706
+  loss_type: ListLoss
+  loss_list:
+  - answer_loss
+  vis_loss_list:
+  - answer_loss

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-20:02:52.972062/config.yaml

@@ -0,0 +1,139 @@
+name: sqa3d_sft_align_run1
+rng_seed: 42
+num_gpu: 4
+mode: train
+note: sqa3d_sft_align_run1
+naming_keywords:
+- dataloader.batchsize
+- task
+- note
+- time
+base_dir: results
+exp_dir: !!python/object/apply:pathlib.PosixPath
+- results
+- sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1
+- 2025-07-02-20:02:52.972062
+save_frequency: 10
+resume: false
+is_pretrain: false
+debug:
+  flag: false
+  hard_debug: false
+  debug_size: 20
+logger:
+  name: wandb
+  entity: yem
+  run_id: 4909845y
+data:
+  train:
+  - ScanNetSQA3D
+  val:
+  - ScanNetSQA3D
+  test:
+  - ScanNetSQA3D
+  args:
+    max_obj_len: 80
+    max_seq_len: 50
+    num_points: 1024
+    pc_type: pred
+    sem_type: '607'
+    filter_lang: false
+    rot_aug: true
+  ScanNetSQA3D:
+    train:
+      use_unanswer: true
+    val:
+      use_unanswer: true
+    test:
+      use_unanswer: true
+  use_voxel: false
+  scan_family_base: existing_datasets/ScanNet
+  rscan_base: existing_datasets/3RScan
+task: SQA3D
+data_wrapper:
+  train: ScanFamilyDatasetWrapperOld
+  val: ScanFamilyDatasetWrapperOld
+  test: ScanFamilyDatasetWrapperOld
+trainer: DefaultTrainer
+ckpt_path: ''
+pretrain_ckpt_path: ''
+dataloader:
+  batchsize: 6
+  num_workers: 2
+  balance_dataset: false
+  filter_empty_annotations: false
+solver:
+  gradient_accumulation_steps: 1
+  epochs_per_save: 10
+  epochs_per_eval: 1
+  lr: 0.0001
+  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: SQA3DEval
+  save: false
+model:
+  name: OpenVocab
+  language:
+    name: BERTLanguageEncoder
+    args:
+      weights: bert-base-uncased
+      hidden_size: 768
+      num_hidden_layers: 4
+      num_attention_heads: 12
+      type_vocab_size: 2
+    lr: 1.0e-05
+  vision:
+    name: qihoo360/fg-clip-base
+    args:
+      backbone: pointnet++
+      hidden_size: 768
+      freeze: true
+      path: pretrained_weights/pointnetpp-open-bert
+      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: 0.0001
+  grounding:
+    name: UnifiedSpatialCrossEncoderV2
+    args:
+      hidden_size: 768
+      num_attention_heads: 12
+      num_layers: 4
+      dim_feedforward: 2048
+      dim_loc: 6
+    lr: 0.0001
+  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: 706
+  loss_type: ListLoss
+  loss_list:
+  - answer_loss
+  vis_loss_list:
+  - answer_loss

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-20:02:52.973200/config.yaml

@@ -0,0 +1,139 @@
+name: sqa3d_sft_align_run1
+rng_seed: 42
+num_gpu: 4
+mode: train
+note: sqa3d_sft_align_run1
+naming_keywords:
+- dataloader.batchsize
+- task
+- note
+- time
+base_dir: results
+exp_dir: !!python/object/apply:pathlib.PosixPath
+- results
+- sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1
+- 2025-07-02-20:02:52.973200
+save_frequency: 10
+resume: false
+is_pretrain: false
+debug:
+  flag: false
+  hard_debug: false
+  debug_size: 20
+logger:
+  name: wandb
+  entity: yem
+  run_id: ec5cs7nr
+data:
+  train:
+  - ScanNetSQA3D
+  val:
+  - ScanNetSQA3D
+  test:
+  - ScanNetSQA3D
+  args:
+    max_obj_len: 80
+    max_seq_len: 50
+    num_points: 1024
+    pc_type: pred
+    sem_type: '607'
+    filter_lang: false
+    rot_aug: true
+  ScanNetSQA3D:
+    train:
+      use_unanswer: true
+    val:
+      use_unanswer: true
+    test:
+      use_unanswer: true
+  use_voxel: false
+  scan_family_base: existing_datasets/ScanNet
+  rscan_base: existing_datasets/3RScan
+task: SQA3D
+data_wrapper:
+  train: ScanFamilyDatasetWrapperOld
+  val: ScanFamilyDatasetWrapperOld
+  test: ScanFamilyDatasetWrapperOld
+trainer: DefaultTrainer
+ckpt_path: ''
+pretrain_ckpt_path: ''
+dataloader:
+  batchsize: 6
+  num_workers: 2
+  balance_dataset: false
+  filter_empty_annotations: false
+solver:
+  gradient_accumulation_steps: 1
+  epochs_per_save: 10
+  epochs_per_eval: 1
+  lr: 0.0001
+  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: SQA3DEval
+  save: false
+model:
+  name: OpenVocab
+  language:
+    name: BERTLanguageEncoder
+    args:
+      weights: bert-base-uncased
+      hidden_size: 768
+      num_hidden_layers: 4
+      num_attention_heads: 12
+      type_vocab_size: 2
+    lr: 1.0e-05
+  vision:
+    name: qihoo360/fg-clip-base
+    args:
+      backbone: pointnet++
+      hidden_size: 768
+      freeze: true
+      path: pretrained_weights/pointnetpp-open-bert
+      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: 0.0001
+  grounding:
+    name: UnifiedSpatialCrossEncoderV2
+    args:
+      hidden_size: 768
+      num_attention_heads: 12
+      num_layers: 4
+      dim_feedforward: 2048
+      dim_loc: 6
+    lr: 0.0001
+  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: 706
+  loss_type: ListLoss
+  loss_list:
+  - answer_loss
+  vis_loss_list:
+  - answer_loss

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-20:02:52.978410/config.yaml

@@ -0,0 +1,139 @@
+name: sqa3d_sft_align_run1
+rng_seed: 42
+num_gpu: 4
+mode: train
+note: sqa3d_sft_align_run1
+naming_keywords:
+- dataloader.batchsize
+- task
+- note
+- time
+base_dir: results
+exp_dir: !!python/object/apply:pathlib.PosixPath
+- results
+- sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1
+- 2025-07-02-20:02:52.978410
+save_frequency: 10
+resume: false
+is_pretrain: false
+debug:
+  flag: false
+  hard_debug: false
+  debug_size: 20
+logger:
+  name: wandb
+  entity: yem
+  run_id: 5l5wypr6
+data:
+  train:
+  - ScanNetSQA3D
+  val:
+  - ScanNetSQA3D
+  test:
+  - ScanNetSQA3D
+  args:
+    max_obj_len: 80
+    max_seq_len: 50
+    num_points: 1024
+    pc_type: pred
+    sem_type: '607'
+    filter_lang: false
+    rot_aug: true
+  ScanNetSQA3D:
+    train:
+      use_unanswer: true
+    val:
+      use_unanswer: true
+    test:
+      use_unanswer: true
+  use_voxel: false
+  scan_family_base: existing_datasets/ScanNet
+  rscan_base: existing_datasets/3RScan
+task: SQA3D
+data_wrapper:
+  train: ScanFamilyDatasetWrapperOld
+  val: ScanFamilyDatasetWrapperOld
+  test: ScanFamilyDatasetWrapperOld
+trainer: DefaultTrainer
+ckpt_path: ''
+pretrain_ckpt_path: ''
+dataloader:
+  batchsize: 6
+  num_workers: 2
+  balance_dataset: false
+  filter_empty_annotations: false
+solver:
+  gradient_accumulation_steps: 1
+  epochs_per_save: 10
+  epochs_per_eval: 1
+  lr: 0.0001
+  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: SQA3DEval
+  save: false
+model:
+  name: OpenVocab
+  language:
+    name: BERTLanguageEncoder
+    args:
+      weights: bert-base-uncased
+      hidden_size: 768
+      num_hidden_layers: 4
+      num_attention_heads: 12
+      type_vocab_size: 2
+    lr: 1.0e-05
+  vision:
+    name: qihoo360/fg-clip-base
+    args:
+      backbone: pointnet++
+      hidden_size: 768
+      freeze: true
+      path: pretrained_weights/pointnetpp-open-bert
+      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: 0.0001
+  grounding:
+    name: UnifiedSpatialCrossEncoderV2
+    args:
+      hidden_size: 768
+      num_attention_heads: 12
+      num_layers: 4
+      dim_feedforward: 2048
+      dim_loc: 6
+    lr: 0.0001
+  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: 706
+  loss_type: ListLoss
+  loss_list:
+  - answer_loss
+  vis_loss_list:
+  - answer_loss

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-20:02:52.979475/config.yaml

@@ -0,0 +1,139 @@
+name: sqa3d_sft_align_run1
+rng_seed: 42
+num_gpu: 4
+mode: train
+note: sqa3d_sft_align_run1
+naming_keywords:
+- dataloader.batchsize
+- task
+- note
+- time
+base_dir: results
+exp_dir: !!python/object/apply:pathlib.PosixPath
+- results
+- sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1
+- 2025-07-02-20:02:52.979475
+save_frequency: 10
+resume: false
+is_pretrain: false
+debug:
+  flag: false
+  hard_debug: false
+  debug_size: 20
+logger:
+  name: wandb
+  entity: yem
+  run_id: czavfly4
+data:
+  train:
+  - ScanNetSQA3D
+  val:
+  - ScanNetSQA3D
+  test:
+  - ScanNetSQA3D
+  args:
+    max_obj_len: 80
+    max_seq_len: 50
+    num_points: 1024
+    pc_type: pred
+    sem_type: '607'
+    filter_lang: false
+    rot_aug: true
+  ScanNetSQA3D:
+    train:
+      use_unanswer: true
+    val:
+      use_unanswer: true
+    test:
+      use_unanswer: true
+  use_voxel: false
+  scan_family_base: existing_datasets/ScanNet
+  rscan_base: existing_datasets/3RScan
+task: SQA3D
+data_wrapper:
+  train: ScanFamilyDatasetWrapperOld
+  val: ScanFamilyDatasetWrapperOld
+  test: ScanFamilyDatasetWrapperOld
+trainer: DefaultTrainer
+ckpt_path: ''
+pretrain_ckpt_path: ''
+dataloader:
+  batchsize: 6
+  num_workers: 2
+  balance_dataset: false
+  filter_empty_annotations: false
+solver:
+  gradient_accumulation_steps: 1
+  epochs_per_save: 10
+  epochs_per_eval: 1
+  lr: 0.0001
+  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: SQA3DEval
+  save: false
+model:
+  name: OpenVocab
+  language:
+    name: BERTLanguageEncoder
+    args:
+      weights: bert-base-uncased
+      hidden_size: 768
+      num_hidden_layers: 4
+      num_attention_heads: 12
+      type_vocab_size: 2
+    lr: 1.0e-05
+  vision:
+    name: qihoo360/fg-clip-base
+    args:
+      backbone: pointnet++
+      hidden_size: 768
+      freeze: true
+      path: pretrained_weights/pointnetpp-open-bert
+      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: 0.0001
+  grounding:
+    name: UnifiedSpatialCrossEncoderV2
+    args:
+      hidden_size: 768
+      num_attention_heads: 12
+      num_layers: 4
+      dim_feedforward: 2048
+      dim_loc: 6
+    lr: 0.0001
+  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: 706
+  loss_type: ListLoss
+  loss_list:
+  - answer_loss
+  vis_loss_list:
+  - answer_loss

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-20:04:42.012763/ckpt/best.pth/custom_checkpoint_0.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:52489de91c416da2004cf429082fd494bce5f4aa047fcf6be9268c88b110e764
+size 1040

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-20:04:42.012763/ckpt/best.pth/model_1.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:60d95b10b6e140a9626a7058d5038528f2ff80148dc4569b881db56052046509
+size 40

results/sqa3d_sft_align_run1_b24_SQA3D_ScanNetSQA3D_sqa3d_sft_align_run1/2025-07-02-20:04:42.012763/ckpt/best.pth/random_states_0.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a7ed0461df701b7cc209c0dd81ee363eb688814c4f8e18a719b4a71ccaf5fc2f
+size 15124