Upload runs/cifar100_weighted_ADAMW_WarmRestart_20251119_200313/config.yaml with huggingface_hub

runs/cifar100_weighted_ADAMW_WarmRestart_20251119_200313/config.yaml

@@ -0,0 +1,57 @@
+adamw_betas:
+- 0.9
+- 0.999
+adamw_eps: 1.0e-08
+batch_size: 128
+beatrix_tau: 0.25
+checkpoint_upload_interval: 25
+cutout_length: 16
+dataset: cifar100
+device: cuda
+drop_path_rate: 0.1
+dropout: 0.1
+embed_dim: 512
+fusion_mode: weighted
+fusion_window: 32
+grad_clip: 1.0
+hf_repo_name: vit-beans-v3
+hf_token: null
+hf_username: AbstractPhil
+image_size: 32
+k_simplex: 4
+label_smoothing: 0.1
+learning_rate: 0.0003
+log_interval: 50
+lr_gamma: 0.2
+lr_milestones:
+- 60
+- 120
+- 160
+min_lr: 1.0e-07
+model_name: vit-beans-v3
+num_classes: 100
+num_epochs: 300
+num_fusion_blocks: 6
+num_heads: 8
+num_workers: 4
+optimizer_type: adamw
+patch_size: 4
+precompute_geometric: true
+restart_mult: 2
+restart_period: 20
+run_name: cifar100_weighted_ADAMW_WarmRestart_20251119_200313
+save_interval: 10
+scheduler_type: cosine_restarts
+seed: 42
+sgd_momentum: 0.9
+sgd_nesterov: true
+upload_to_hf: true
+use_augmentation: true
+use_autoaugment: true
+use_beatrix: true
+use_cutout: false
+use_mixed_precision: false
+use_torch_compile: true
+warmup_epochs: 10
+weight_decay: 0.05
+weights_dir: weights