first commit

datt_scannet/wandb/latest-run/files/config.yaml

@@ -0,0 +1,262 @@
+wandb_version: 1
+
+CONFIG:
+  desc: null
+  value: datt_mix_ctrs.yaml
+MODEL:
+  desc: null
+  value:
+    TYPE: dat
+    NAME: dat_plus_plus
+    PRETRAINED: pretrained_weights/dat_pp_tiny_in1k_224.pth
+    DAT:
+      img_size: 224
+      patch_size: 4
+      num_classes: 1000
+      expansion: 4
+      dim_stem: 64
+      dims:
+      - 64
+      - 128
+      - 256
+      - 512
+      depths:
+      - 2
+      - 4
+      - 18
+      - 2
+      stage_spec:
+      - - N
+        - D
+      - - N
+        - D
+        - N
+        - D
+      - - N
+        - D
+        - N
+        - D
+        - N
+        - D
+        - N
+        - D
+        - N
+        - D
+        - N
+        - D
+        - N
+        - D
+        - N
+        - D
+        - N
+        - D
+      - - D
+        - D
+      heads:
+      - 2
+      - 4
+      - 8
+      - 16
+      window_sizes:
+      - 7
+      - 7
+      - 7
+      - 7
+      groups:
+      - 1
+      - 2
+      - 4
+      - 8
+      use_pes:
+      - true
+      - true
+      - true
+      - true
+      dwc_pes:
+      - false
+      - false
+      - false
+      - false
+      strides:
+      - 8
+      - 4
+      - 2
+      - 1
+      offset_range_factor:
+      - -1
+      - -1
+      - -1
+      - -1
+      no_offs:
+      - false
+      - false
+      - false
+      - false
+      fixed_pes:
+      - false
+      - false
+      - false
+      - false
+      use_dwc_mlps:
+      - true
+      - true
+      - true
+      - true
+      use_lpus:
+      - true
+      - true
+      - true
+      - true
+      use_conv_patches: true
+      ksizes:
+      - 9
+      - 7
+      - 5
+      - 3
+      nat_ksizes:
+      - 7
+      - 7
+      - 7
+      - 7
+      drop_rate: 0.0
+      attn_drop_rate: 0.0
+      drop_path_rate: 0.2
+      clip_dim: 1024
+      quantization:
+        method: jacob
+        weight:
+          num_bits: 8
+          axis: per_channel
+          calib_method: max
+        activation:
+          num_bits: 8
+          axis: per_tensor
+          calib_method: max
+        calibration:
+          num_batch: 2
+          modal: rgbd
+TRAIN:
+  desc: null
+  value:
+    EPOCHS: 120
+    WARMUP_EPOCHS: 0
+    BASE_LR: 1.0e-06
+    WEIGHT_DECAY: 0.05
+    WARMUP_LR: 5.0e-07
+    MIN_LR: 5.0e-07
+    CLIP_GRAD: 5.0
+    LR_SCHEDULER:
+      NAME: cosine
+      DECAY_EPOCHS: 30
+      DECAY_RATE: 0.1
+    OPTIMIZER:
+      NAME: adamw
+      EPS: 1.0e-08
+      BETAS:
+      - 0.9
+      - 0.999
+      MOMENTUM: 0.9
+    CRETERION:
+      NAME: TripletMarginLoss
+      TRIPLET_MARGIN: 0.3
+      MINING_METHOD:
+        POSITIVE: easy
+        NEGATIVE: random
+        TOTAL: semihard
+      NEG_NUM: 3
+      KD:
+        ENABLE: false
+        WEIGHT: 10
+      WEIGHT: 0.05
+DATA:
+  desc: null
+  value:
+    DATASET: scannet
+    ROOT: dbs
+    BATCH_SIZE: 4
+    VAL_BATCH_SIZE: 32
+    EMBEDDING_BATCH_SIZE: 32
+    DEPTH_TRANSFORM: rgb
+    IS_SUBSET: true
+    LABEL_TYPE: pseudo_labels
+    DATASET_THRESHOLD: 0.25
+GPU:
+  desc: null
+  value: 0
+SEED:
+  desc: null
+  value: 1
+CLIP_MODEL:
+  desc: null
+  value:
+    NAME: ViT-g-14
+    PRETRAINED: laion2b_s34b_b88k
+MODAL:
+  desc: null
+  value: depth
+PHASE:
+  desc: null
+  value: train_ctrs
+CKPT:
+  desc: null
+  value: logs/datt_0215_215348/wandb/latest-run/files/src/best_model.pth
+MIX_INPUT:
+  desc: null
+  value:
+    ENABLE: true
+    NUM_MIX: 2
+    WEIGHTED: false
+ATTENTION:
+  desc: null
+  value:
+    ENABLE: false
+WEIGHTED_SAMPLE:
+  desc: null
+  value:
+    ENABLE: false
+cmd:
+  desc: null
+  value: run.py --phase=train_ctrs --config=configs/scannet/datt_mix_ctrs.yaml --quant_config=quantization_configs/d2_jacob.yaml
+_wandb:
+  desc: null
+  value:
+    code_path: code/run.py
+    python_version: 3.10.13
+    cli_version: 0.16.3
+    framework: huggingface
+    huggingface_version: 4.37.2
+    is_jupyter_run: false
+    is_kaggle_kernel: false
+    start_time: 1708266273.066213
+    t:
+      1:
+      - 1
+      - 5
+      - 11
+      - 41
+      - 49
+      - 53
+      - 55
+      - 63
+      - 80
+      2:
+      - 1
+      - 5
+      - 11
+      - 41
+      - 49
+      - 53
+      - 55
+      - 63
+      - 80
+      3:
+      - 3
+      - 13
+      - 16
+      - 23
+      4: 3.10.13
+      5: 0.16.3
+      6: 4.37.2
+      8:
+      - 5
+      13: linux-x86_64

datt_scannet/wandb/latest-run/files/src/best_model.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8dc2a4016a949a2a9cad3ffd68eccbb99b3d5e389d67896ee40e4671a7a4a837
+size 98227739

datt_scannet/wandb/latest-run/files/src/datt_mix_ctrs.yaml

@@ -0,0 +1,100 @@
+CONFIG: datt_mix_ctrs.yaml
+MODEL:
+  TYPE: dat
+  NAME: dat_plus_plus
+  PRETRAINED: "pretrained_weights/dat_pp_tiny_in1k_224.pth"
+  DAT:
+    img_size: 224
+    patch_size: 4
+    num_classes: 1000
+    expansion: 4
+    dim_stem: 64
+    dims: [64, 128, 256, 512]
+    depths: [2, 4, 18, 2]
+    stage_spec: [[N, D], [N, D, N, D], [N, D, N, D, N, D, N, D, N, D, N, D, N, D, N, D, N, D], [D, D]]
+    heads: [2, 4, 8, 16]
+    window_sizes: [7, 7, 7, 7]
+    groups: [1, 2, 4, 8]
+    use_pes: [True, True, True, True]
+    dwc_pes: [False, False, False, False]
+    strides: [8, 4, 2, 1]
+    offset_range_factor: [-1, -1, -1, -1]
+    no_offs: [False, False, False, False]
+    fixed_pes: [False, False, False, False]
+    use_dwc_mlps: [True, True, True, True]
+    use_lpus: [True, True, True, True]
+    use_conv_patches: True
+    ksizes: [9, 7, 5, 3]
+    nat_ksizes: [7, 7, 7, 7]
+    drop_rate: 0.0
+    attn_drop_rate: 0.0
+    drop_path_rate: 0.2
+    # quant_bit: 8
+    clip_dim: 1024
+
+
+TRAIN:
+  EPOCHS: 120
+  WARMUP_EPOCHS: 0
+  BASE_LR: 1e-6
+  WEIGHT_DECAY: 0.05
+  WARMUP_LR: 5e-7
+  MIN_LR: 5e-7
+  CLIP_GRAD: 5.0
+  LR_SCHEDULER:
+    NAME: "cosine"
+    DECAY_EPOCHS: 30
+    DECAY_RATE: 0.1
+  OPTIMIZER:
+    NAME: "adamw"
+    EPS: 1e-8
+    BETAS: [0.9, 0.999]
+    MOMENTUM: 0.9
+  CRETERION:
+    NAME: "TripletMarginLoss"  # MSE  
+    TRIPLET_MARGIN: 0.3
+    MINING_METHOD:
+      POSITIVE: "easy"  # easy, hard, random
+      NEGATIVE: "random"  # easy, hard, random
+      TOTAL: "semihard"   # all, semihard, hard
+    NEG_NUM: 3
+    KD:
+      ENABLE: False
+      WEIGHT: 10
+    WEIGHT: 5e-2
+    
+DATA:
+  DATASET: "scannet"  
+  ROOT: "dbs"
+  BATCH_SIZE: 4
+  VAL_BATCH_SIZE: 32
+  EMBEDDING_BATCH_SIZE: 32
+  DEPTH_TRANSFORM: "rgb"
+  IS_SUBSET: True
+  LABEL_TYPE: "pseudo_labels"  # "gt" or "clip_vitb32"
+  DATASET_THRESHOLD: 0.25
+
+
+GPU: 0
+SEED: 1
+
+CLIP_MODEL: 
+  NAME: "ViT-g-14"      # "ViT-B/32"
+  PRETRAINED: "laion2b_s34b_b88k"
+MODAL: "depth"  # "rgb" or "depth"
+PHASE: "train_ctrs"  # "train_ctrs" # "test"
+CKPT: "logs/datt_0215_215348/wandb/latest-run/files/src/best_model.pth"
+
+# [] Mix Input
+MIX_INPUT:
+  ENABLE: True
+  NUM_MIX: 2
+  WEIGHTED: False
+  
+# [] Attention
+ATTENTION:
+  ENABLE: False
+
+# [] Weighted samples
+WEIGHTED_SAMPLE:
+  ENABLE: False