关于在UCF101数据集上训练的问题

[xiezexun]

作者你好，看到你的论文和模型，我深受启发。我想在UCF101数据集上训练，但是遇到了以下问题：配置文件修改哪些内容？直接复制kenitecs400的配置文件可以吗？除了配置文件其他Python代码部分是否需要修改呢？

[Andy1621]

复制配置文件可以的，但是你需要调整相应的训练参数，python文件不需要修改，可以直接使用kinetics的数据集文件

[xiezexun]

感谢您的回复，我直接复制配置文件之后，修改了训练参数，但是训练ucf101的效果很不好，loss从头到尾一直在5.0左右波动，验证集的准确率也只有1%不到，不知道是哪里出了问题，恳请您给予解答。配置文件内容如下： TRAIN: ENABLE: True DATASET: kinetics BATCH_SIZE: 2 EVAL_PERIOD: 1 CHECKPOINT_PERIOD: 10 AUTO_RESUME: True DATA: USE_OFFSET_SAMPLING: True DECODING_BACKEND: decord NUM_FRAMES: 16 SAMPLING_RATE: 4 TRAIN_JITTER_SCALES: [256, 320] TRAIN_CROP_SIZE: 224 TEST_CROP_SIZE: 224 INPUT_CHANNEL_NUM: [3] PATH_PREFIX: ../../data/ucf101/videos PATH_TO_DATA_DIR: ../data_list/ucf101/split1 TRAIN_JITTER_SCALES_RELATIVE: [0.08, 1.0] TRAIN_JITTER_ASPECT_RELATIVE: [0.75, 1.3333] UNIFORMERV2: BACKBONE: 'uniformerv2_l14' N_LAYERS: 4 N_DIM: 1024 N_HEAD: 16 MLP_FACTOR: 4.0 BACKBONE_DROP_PATH_RATE: 0. DROP_PATH_RATE: 0. MLP_DROPOUT: [0.5, 0.5, 0.5, 0.5] CLS_DROPOUT: 0.5 RETURN_LIST: [20, 21, 22, 23] NO_LMHRA: True TEMPORAL_DOWNSAMPLE: False AUG: NUM_SAMPLE: 1 ENABLE: True COLOR_JITTER: 0.4 AA_TYPE: rand-m7-n4-mstd0.5-inc1 INTERPOLATION: bicubic RE_PROB: 0. RE_MODE: pixel RE_COUNT: 1 RE_SPLIT: False BN: USE_PRECISE_STATS: False NUM_BATCHES_PRECISE: 200 SOLVER: ZERO_WD_1D_PARAM: True BASE_LR_SCALE_NUM_SHARDS: True BASE_LR: 4e-4 COSINE_AFTER_WARMUP: True COSINE_END_LR: 1e-6 WARMUP_START_LR: 1e-6 WARMUP_EPOCHS: 0. LR_POLICY: cosine MAX_EPOCH: 50 MOMENTUM: 0.9 WEIGHT_DECAY: 0.05 OPTIMIZING_METHOD: adamw MODEL: NUM_CLASSES: 101 ARCH: uniformerv2 MODEL_NAME: Uniformerv2 LOSS_FUNC: cross_entropy DROPOUT_RATE: 0.5 USE_CHECKPOINT: False CHECKPOINT_NUM: [0] TEST: ENABLE: True DATASET: kinetics_sparse BATCH_SIZE: 8 NUM_SPATIAL_CROPS: 1 NUM_ENSEMBLE_VIEWS: 1 DATA_LOADER: NUM_WORKERS: 8 PIN_MEMORY: True TENSORBOARD: ENABLE: False NUM_GPUS: 1 NUM_SHARDS: 1 RNG_SEED: 0 OUTPUT_DIR: ../exp/ucf101

[Andy1621]

不知道是否解决？