rwightman HF staff commited on
Commit
b152001
1 Parent(s): e6e5065

Upload train_hparams.yaml with huggingface_hub

Browse files
Files changed (1) hide show
  1. train_hparams.yaml +134 -0
train_hparams.yaml ADDED
@@ -0,0 +1,134 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ aa: rand-m7-inc1-mstd1.0-n3
2
+ amp: true
3
+ amp_dtype: float16
4
+ amp_impl: native
5
+ aug_repeats: 3.0
6
+ aug_splits: 0
7
+ batch_size: 256
8
+ bce_loss: false
9
+ bce_target_thresh: null
10
+ bn_eps: null
11
+ bn_momentum: null
12
+ channels_last: false
13
+ checkpoint_hist: 10
14
+ class_map: ''
15
+ clip_grad: 5.0
16
+ clip_mode: norm
17
+ color_jitter: null
18
+ cooldown_epochs: 10
19
+ crop_pct: 0.95
20
+ cutmix: 1.0
21
+ cutmix_minmax: null
22
+ data: null
23
+ data_dir: /data/imagenet/
24
+ dataset: imagenet
25
+ dataset_download: false
26
+ decay_epochs: 100
27
+ decay_milestones:
28
+ - 90
29
+ - 180
30
+ - 270
31
+ decay_rate: 0.1
32
+ dist_bn: reduce
33
+ drop: 0.0
34
+ drop_block: null
35
+ drop_connect: null
36
+ drop_path: 0.2
37
+ epoch_repeats: 0.0
38
+ epochs: 600
39
+ eval_metric: top1
40
+ experiment: ''
41
+ fast_norm: false
42
+ force_cpu: false
43
+ fuser: ''
44
+ gp: null
45
+ grad_accum_steps: 1
46
+ grad_checkpointing: false
47
+ head_init_bias: null
48
+ head_init_scale: 0.0
49
+ hflip: 0.5
50
+ img_size: 256
51
+ in_chans: null
52
+ initial_checkpoint: ''
53
+ input_size: null
54
+ interpolation: ''
55
+ jsd_loss: false
56
+ layer_decay: null
57
+ local_rank: 0
58
+ log_interval: 50
59
+ log_wandb: false
60
+ lr: null
61
+ lr_base: 0.0008
62
+ lr_base_scale: ''
63
+ lr_base_size: 512
64
+ lr_cycle_decay: 0.5
65
+ lr_cycle_limit: 1
66
+ lr_cycle_mul: 1.0
67
+ lr_k_decay: 1.0
68
+ lr_noise: null
69
+ lr_noise_pct: 0.67
70
+ lr_noise_std: 1.0
71
+ mean: null
72
+ min_lr: 5.0e-07
73
+ mixup: 0.8
74
+ mixup_mode: batch
75
+ mixup_off_epoch: 0
76
+ mixup_prob: 1.0
77
+ mixup_switch_prob: 0.5
78
+ model: vit_betwixt_patch16_reg1_gap_256
79
+ model_ema: false
80
+ model_ema_decay: 0.9998
81
+ model_ema_force_cpu: false
82
+ model_kwargs:
83
+ fix_init: true
84
+ momentum: 0.9
85
+ no_aug: false
86
+ no_ddp_bb: false
87
+ no_prefetcher: false
88
+ no_resume_opt: false
89
+ num_classes: null
90
+ opt: nadamw
91
+ opt_betas: null
92
+ opt_eps: 1.0e-08
93
+ opt_kwargs: {}
94
+ output: ''
95
+ patience_epochs: 10
96
+ pin_mem: true
97
+ pretrained: false
98
+ ratio:
99
+ - 0.75
100
+ - 1.3333333333333333
101
+ recount: 1
102
+ recovery_interval: 0
103
+ remode: pixel
104
+ reprob: 0.2
105
+ resplit: false
106
+ resume: ''
107
+ save_images: false
108
+ scale:
109
+ - 0.08
110
+ - 1.0
111
+ sched: cosine
112
+ sched_on_updates: true
113
+ seed: 0
114
+ smoothing: 0.1
115
+ split_bn: false
116
+ start_epoch: null
117
+ std: null
118
+ sync_bn: false
119
+ synchronize_step: false
120
+ torchcompile: inductor
121
+ torchscript: false
122
+ train_interpolation: random
123
+ train_split: train
124
+ tta: 0
125
+ use_multi_epochs_loader: false
126
+ val_split: validation
127
+ validation_batch_size: null
128
+ vflip: 0.0
129
+ warmup_epochs: 20
130
+ warmup_lr: 5.0e-07
131
+ warmup_prefix: false
132
+ weight_decay: 0.06
133
+ worker_seeding: all
134
+ workers: 8