rwightman HF staff commited on
Commit
bb463c9
1 Parent(s): a403173

Upload train_hparams.yaml with huggingface_hub

Browse files
Files changed (1) hide show
  1. train_hparams.yaml +149 -0
train_hparams.yaml ADDED
@@ -0,0 +1,149 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ aa: rand-m7-inc1-mstd1.0-n3
2
+ amp: true
3
+ amp_dtype: float16
4
+ amp_impl: native
5
+ aug_repeats: 0.0
6
+ aug_splits: 0
7
+ batch_size: 256
8
+ bce_loss: true
9
+ bce_pos_weight: null
10
+ bce_sum: true
11
+ bce_target_thresh: 0.2
12
+ bn_eps: null
13
+ bn_momentum: null
14
+ channels_last: false
15
+ checkpoint_hist: 10
16
+ class_map: ''
17
+ clip_grad: 5.0
18
+ clip_mode: norm
19
+ color_jitter: null
20
+ color_jitter_prob: 0.2
21
+ cooldown_epochs: 10
22
+ crop_pct: 0.95
23
+ cutmix: 1.0
24
+ cutmix_minmax: null
25
+ data: null
26
+ data_dir: /data/tfds/
27
+ dataset: tfds/imagenet12k
28
+ dataset_download: false
29
+ decay_epochs: 100
30
+ decay_milestones:
31
+ - 90
32
+ - 180
33
+ - 270
34
+ decay_rate: 0.1
35
+ device: cuda
36
+ device_modules: null
37
+ dist_bn: reduce
38
+ drop: 0.0
39
+ drop_block: null
40
+ drop_connect: null
41
+ drop_path: 0.2
42
+ epoch_repeats: 0.0
43
+ epochs: 150
44
+ eval_metric: top1
45
+ experiment: ''
46
+ fast_norm: false
47
+ force_cpu: false
48
+ fuser: ''
49
+ gaussian_blur_prob: 0.05
50
+ gp: null
51
+ grad_accum_steps: 1
52
+ grad_checkpointing: false
53
+ grayscale_prob: 0.1
54
+ head_init_bias: null
55
+ head_init_scale: 0.0
56
+ hflip: 0.5
57
+ img_size: 256
58
+ in_chans: null
59
+ initial_checkpoint: ''
60
+ input_img_mode: null
61
+ input_key: null
62
+ input_size: null
63
+ interpolation: ''
64
+ jsd_loss: false
65
+ layer_decay: null
66
+ local_rank: 0
67
+ log_interval: 50
68
+ log_wandb: false
69
+ lr: null
70
+ lr_base: 0.0001
71
+ lr_base_scale: ''
72
+ lr_base_size: 512
73
+ lr_cycle_decay: 0.5
74
+ lr_cycle_limit: 1
75
+ lr_cycle_mul: 1.0
76
+ lr_k_decay: 1.0
77
+ lr_noise: null
78
+ lr_noise_pct: 0.67
79
+ lr_noise_std: 1.0
80
+ mean: null
81
+ min_lr: 5.0e-07
82
+ mixup: 0.2
83
+ mixup_mode: batch
84
+ mixup_off_epoch: 0
85
+ mixup_prob: 1.0
86
+ mixup_switch_prob: 0.5
87
+ model: vit_mediumd_patch16_reg4_gap_256
88
+ model_ema: true
89
+ model_ema_decay: 0.99993
90
+ model_ema_force_cpu: false
91
+ model_ema_warmup: false
92
+ model_kwargs:
93
+ fix_init: true
94
+ momentum: 0.9
95
+ no_aug: false
96
+ no_ddp_bb: false
97
+ no_prefetcher: false
98
+ no_resume_opt: false
99
+ num_classes: 11821
100
+ opt: nadamw
101
+ opt_betas: null
102
+ opt_eps: 1.0e-08
103
+ opt_kwargs: {}
104
+ output: ''
105
+ patience_epochs: 10
106
+ pin_mem: true
107
+ pretrained: false
108
+ pretrained_path: null
109
+ ratio:
110
+ - 0.75
111
+ - 1.3333333333333333
112
+ recount: 1
113
+ recovery_interval: 0
114
+ remode: pixel
115
+ reprob: 0.2
116
+ resplit: false
117
+ resume: ''
118
+ save_images: false
119
+ scale:
120
+ - 0.2
121
+ - 1.0
122
+ sched: cosine
123
+ sched_on_updates: true
124
+ seed: 0
125
+ smoothing: 0.0
126
+ split_bn: false
127
+ start_epoch: null
128
+ std: null
129
+ sync_bn: false
130
+ synchronize_step: false
131
+ target_key: null
132
+ torchcompile: inductor
133
+ torchscript: false
134
+ train_crop_mode: rkrr
135
+ train_interpolation: random
136
+ train_num_samples: null
137
+ train_split: train
138
+ tta: 0
139
+ use_multi_epochs_loader: false
140
+ val_num_samples: null
141
+ val_split: validation
142
+ validation_batch_size: null
143
+ vflip: 0.0
144
+ warmup_epochs: 3
145
+ warmup_lr: 5.0e-07
146
+ warmup_prefix: false
147
+ weight_decay: 0.2
148
+ worker_seeding: all
149
+ workers: 8