single-shot-robot

Josselin's group workspace

Group: 3-DICE-1-Skil-Product

1-20

of 20

Tags

Notes

Author

josselin

State

Finished

Start time

September 9th, 2023 1:21:41 AM

Runtime

42m 2s

Tracked hours

41m 56s

Run path

single-shot-robot/SKIL-Reachbot_vessels_combined_segmentation_prod/cyx5sqof

Linux-5.15.0-79-generic-x86_64-with-glibc2.17

Python version

3.8.16

Git repository

git clone git@github.com:JosselinSomervilleRoberts/ReachBot-Predictor.git

Git state

git checkout -b "3-DICE-1-Skil-Product_5WETWABTQP" 745ad634788c4f3adb045507955d6ebbbf6b6c32

Command

tools/train.py paper_configs/vit/vessels/skil_prod.py --amp

System Hardware

CPU count	18
Logical CPU count	36
GPU count	1
GPU type	NVIDIA RTX A6000

W&B CLI Version

0.15.4

Group

3-DICE-1-Skil-Product

Config parameters are your model's inputs. Learn more

▶
Config parameters:{} 54 keys
- CHECKPOINT:
  "pretrain/upernet_vit-b16_mln_512x512_80k_ade20k_20210624_130547-0403cee1_fix.pth"
- ▶
  crop_size:[] 2 items
  - 0:
    512
  - 1:
    512
- ▶
  data_preprocessor:{} 7 keys
- data_root:
  "../datasets/vessels/combined"
- DATASET:
  "vessels_combined"
- dataset_type:
  "ReachbotDataset"
- ▶
  datasets_mappings:{} 6 keys
- DEBUG_PATH:
  null
- DEBUG_STEP_INTERVAL:
  -1
- ▶
  default_checkpoint:{} 2 keys
  - vit_vit-b16_mln_upernet_cracks:
    "pretrain/upernet_vit-b16_mln_512x512_80k_ade20k_20210624_130547-0403cee1_fix.pth"
  - vit_vit-b16_mln_upernet_vessels:
    "pretrain/upernet_vit-b16_mln_512x512_80k_ade20k_20210624_130547-0403cee1_fix.pth"
- ▶
  default_hooks:{} 4 keys
  - ▶
    logger:{} 3 keys
    - interval:
      5
    - log_metric_by_epoch:
      false
    - type:
      "LoggerHook"
  - ▶
    param_scheduler:{} 1 key
    - type:
      "ParamSchedulerHook"
  - ▶
    sampler_seed:{} 1 key
    - type:
      "DistSamplerSeedHook"
  - ▶
    timer:{} 1 key
    - type:
      "IterTimerHook"
- default_scope:
  "mmseg"
- ▶
  env_cfg:{} 3 keys
  - cudnn_benchmark:
    true
  - ▶
    dist_cfg:{} 1 key
    - backend:
      "nccl"
  - ▶
    mp_cfg:{} 2 keys
    - mp_start_method:
      "fork"
    - opencv_num_threads:
      0
- EVAL_EPOCH_INTERVAL:
  5
- launcher:
  "none"
- load_from:
  null
- log_level:
  "INFO"
- ▶
  log_processor:{} 1 key
  - by_epoch:
    false
- LOG_STEP_INTERVAL:
  5
- ▶
  LOSSES:{} 2 keys
  - dice:
    3
  - ▶
    skil_product:[] 2 items
    - 0:
      1
    - ▶
      1:{} 4 keys
      - border_factor:
        0.82
      - border_size:
        20
      - epsilon:
        0.01
      - iterations:
        50
- ▶
  losses_mappings:{} 5 keys
- LR:
  "aggressive"
- memory_available:
  47.0091552734375
- METRIC:
  "ReachbotMetric"
- MODEL:
  "vit_vit-b16_mln_upernet_vessels"
- ▶
  model:{} 9 keys
- ▶
  norm_cfg:{} 2 keys
  - requires_grad:
    true
  - type:
    "SyncBN"
- num_epochs:
  200
- num_training_samples:
  85
- ▶
  optim_wrapper:{} 4 keys
  - loss_scale:
    "dynamic"
  - ▶
    optimizer:{} 4 keys
    - ▶
      betas:[] 2 items
      - 0:
        0.9
      - 1:
        0.999
    - lr:
      0.00005
    - type:
      "AdamW"
    - weight_decay:
      0.01
  - ▶
    paramwise_cfg:{} 1 key
    - ▶
      custom_keys:{} 3 keys
      - ▶
        cls_token:{} 1 key
        decay_mult:
        0
      - ▶
        norm:{} 1 key
        decay_mult:
        0
      - ▶
        pos_embed:{} 1 key
        decay_mult:
        0
  - type:
    "AmpOptimWrapper"
- ▶
  param_scheduler:[] 2 items
  - ▶
    0:{} 5 keys
  - ▶
    1:{} 6 keys
- resume:
  false
- SAVE_EPOCH_INTERVAL:
  -1
- ▶
  schedules_mappings:{} 2 keys
  - 40k:
    "custom_configs.schedules.40k"
  - aggressive:
    "custom_configs.schedules.aggressive"
- tag:
  "5WETWABTQP"
- ▶
  test_cfg:{} 1 key
  - type:
    "TestLoop"
- ▶
  test_dataloader:{} 5 keys
- ▶
  test_evaluator:{} 1 key
  - type:
    "ReachbotMetric"
- ▶
  test_pipeline:[] 4 items
  - ▶
    0:{} 1 key
    - type:
      "LoadImageFromFile"
  - ▶
    1:{} 3 keys
    - keep_ratio:
      true
    - ▶
      scale:[] 2 items
      - 0:
        1,024
      - 1:
        512
    - type:
      "Resize"
  - ▶
    2:{} 2 keys
    - reduce_zero_label:
      false
    - type:
      "LoadAnnotations"
  - ▶
    3:{} 1 key
    - type:
      "PackSegInputs"
- train_batch_size:
  8
- ▶
  train_cfg:{} 3 keys
  - max_iters:
    2,125
  - type:
    "IterBasedTrainLoop"
  - val_interval:
    53
- ▶
  train_dataloader:{} 5 keys
- ▶
  train_pipeline:[] 7 items
- ▶
  tta_model:{} 1 key
  - type:
    "SegTTAModel"
- USE_WANDB:
  true
- ▶
  val_cfg:{} 1 key
  - type:
    "ValLoop"
- ▶
  visualizer:{} 4 keys
  - alpha:
    0.6
  - name:
    "visualizer"
  - type:
    "SegLocalVisualizer"
  - ▶
    vis_backends:[] 2 items
    - ▶
      0:{} 1 key
      - type:
        "LocalVisBackend"
    - ▶
      1:{} 2 keys
      - ▶
        init_kwargs:{} 4 keys
        entity:
        "single-shot-robot"
        group:
        "3-DICE-1-Skil-Product"
        name:
        "3-DICE-1-Skil-Product_5WETWABTQP"
        project:
        "vessels_combined_segmentation_prod"
      - type:
        "WandbVisBackend"
- WANDB_PROJECT_SUFFIX:
  "_prod"
- WANDB_RUN_NAME:
  "3-DICE-1-Skil-Product"
- work_dir:
  "./work_dirs/skil_prod"

Summary metrics are your model's outputs. Learn more

▶
Summary metrics:{} 18 keys
- aux.acc_seg:
  87.9154052734375
- aux.loss_dice:
  0.421684181690216
- aux.loss_skil_product:
  0.3178885132074356
- avg:
  0.6612195885717914
- data_time:
  0.02217686176300049
- decode.acc_seg:
  94.2534637451172
- decode.loss_dice:
  0.6378354489803314
- decode.loss_skil_product:
  0.2187287375330925
- dice:
  0.391108591634919
- iou:
  0.2585339669818646
- iter:
  2,125
- length:
  0.7102355639190829
- line_distance:
  0.8031418003672446
- loss:
  1.5961369037628177
- lr:
  0
- memory:
  37,775
- time:
  0.9641390800476074
- width:
  0.6682954092221948

This run consumed these artifacts as inputs. Learn more

job

job-git_github.com_JosselinSomervilleRoberts_ReachBot-Predictor.git_mmsegmentation_tools_train.py:v16

This run produced these artifacts as outputs. Total: 2. Learn more

code

source-vessels_combined_segmentation_prod-mmsegmentation_tools_train.py:v111

wandb-history

run-cyx5sqof-history:v0