Upload config.yaml

config.yaml

@@ -0,0 +1,121 @@
+TRAIN:
+  ENABLE: True
+  DATASET: charades
+  BATCH_SIZE: 5
+  EVAL_PERIOD: 5
+  CHECKPOINT_PERIOD: 5
+  AUTO_RESUME: True
+  # CHECKPOINT_FILE_PATH:
+  CHECKPOINT_TYPE: pytorch
+  CHECKPOINT_INFLATE: False
+  MIXED_PRECISION: True
+
+TEST:
+  ENABLE: True
+  DATASET: charades
+  BATCH_SIZE: 5
+  NUM_ENSEMBLE_VIEWS: 2
+  NUM_SPATIAL_CROPS: 1
+  CHECKPOINT_TYPE: pytorch
+  SAVE_RESULTS_PATH: "/mnt/sdb1/Documents/Animals/SlowFast/logs/x3d-l-BaboonLand/results.txt"
+  CHECKPOINT_FILE_PATH: /mnt/sdb1/Documents/Animals/SlowFast/logs/x3d-l-BaboonLand/checkpoints/checkpoint_epoch_00080.pyth
+
+DATA:
+  NUM_FRAMES: 16
+  SAMPLING_RATE: 5
+  TRAIN_JITTER_SCALES: [300, 400]
+  TRAIN_CROP_SIZE: 300
+  TEST_CROP_SIZE: 300
+  TRAIN_CROP_NUM_TEMPORAL: 1
+  INPUT_CHANNEL_NUM: [3]
+  MULTI_LABEL: True
+  RANDOM_FLIP: True
+  SSL_COLOR_JITTER: True
+  SSL_COLOR_BRI_CON_SAT: [0.2, 0.2, 0.2]
+  INV_UNIFORM_SAMPLE: True
+  ENSEMBLE_METHOD: max
+  REVERSE_INPUT_CHANNEL: True
+  PATH_TO_DATA_DIR: "/mnt/NVME/BaboonLand/charades/annotation"
+  PATH_PREFIX: "/mnt/NVME/BaboonLand/charades/dataset/image"
+  DECODING_BACKEND: torchvision
+
+X3D:
+  WIDTH_FACTOR: 2.0
+  DEPTH_FACTOR: 5.0
+  BOTTLENECK_FACTOR: 2.25
+  DIM_C5: 2048
+  DIM_C1: 12
+
+RESNET:
+  ZERO_INIT_FINAL_BN: True
+  TRANS_FUNC: x3d_transform
+  STRIDE_1X1: False
+
+BN:
+  USE_PRECISE_STATS: True
+  NUM_BATCHES_PRECISE: 200
+  NORM_TYPE: sync_batchnorm
+  NUM_SYNC_DEVICES: 1
+  WEIGHT_DECAY: 0.0
+
+SOLVER:
+  BASE_LR: 0.05
+  BASE_LR_SCALE_NUM_SHARDS: True
+  MAX_EPOCH: 120
+  LR_POLICY: cosine
+  WEIGHT_DECAY: 5e-5
+  WARMUP_EPOCHS: 35.0
+  WARMUP_START_LR: 0.01
+  OPTIMIZING_METHOD: sgd
+
+MODEL:
+  NUM_CLASSES: 13
+  ARCH: x3d
+  MODEL_NAME: X3D
+  LOSS_FUNC: EQL
+  HEAD_ACT: sigmoid
+  DROPOUT_RATE: 0.5
+
+DATA_LOADER:
+  NUM_WORKERS: 8
+  PIN_MEMORY: True
+
+NUM_GPUS: 1
+NUM_SHARDS: 1
+RNG_SEED: 0
+OUTPUT_DIR: ./logs/x3d-l-BaboonLand
+LOG_MODEL_INFO: True
+
+TENSORBOARD:
+  ENABLE: True
+  CLASS_NAMES_PATH: /mnt/NVME/BaboonLand/charades/annotation/classes.json
+  CONFUSION_MATRIX:
+    ENABLE: True
+    SUBSET_PATH: /mnt/NVME/BaboonLand/charades/annotation/classes.txt
+  HISTOGRAM:
+    ENABLE: True
+    TOPK: 3
+    SUBSET_PATH: /mnt/NVME/BaboonLand/charades/annotation/classes.txt
+  MODEL_VIS:
+    ENABLE: True
+    MODEL_WEIGHTS: True
+    ACTIVATIONS: True
+    INPUT_VIDEO: True
+    LAYER_LIST: ['s5/pathway0_res14']
+    GRAD_CAM:
+      ENABLE: True
+      LAYER_LIST: ['s5/pathway0_res14']
+
+DEMO:
+  ENABLE: True
+  LABEL_FILE_PATH: /mnt/NVME/BaboonLand/charades/annotation/classes.json
+  INPUT_VIDEO: /mnt/NVME/BaboonLand/charades/dataset/video/B0093.mp4
+  OUTPUT_FILE: /mnt/NVME/BaboonLand/charades/dataset/predict/B0093.mp4
+  THREAD_ENABLE: False
+  THREAD_ENABLE: False
+  NUM_VIS_INSTANCES: 1
+  NUM_CLIPS_SKIP: 1
+  # VIS_MODE: "top-k"
+  # UNCOMMON_CLASS_THRES: 0.3
+  # COMMON_CLASS_THRES: 0.7
+  # COMMON_CLASS_NAMES: ["Graze"]