kaitongg commited on
Commit
ed90fd2
·
verified ·
1 Parent(s): 513630a

Upload folder using huggingface_hub

Browse files
assets.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"learner_class": "BaseLearner", "column_types": {"image": "image_path", "label": "categorical"}, "label_column": "label", "problem_type": "binary", "presets": "medium_quality", "eval_metric_name": "accuracy", "validation_metric_name": "accuracy", "minmax_mode": "max", "output_shape": 2, "save_path": "/content/autogluon_sign_artifacts/predictor", "pretrained": true, "pretrained_path": null, "fit_called": true, "best_score": null, "total_train_time": null, "version": "1.4.0"}
config.yaml ADDED
@@ -0,0 +1,142 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ model:
2
+ names:
3
+ - timm_image
4
+ timm_image:
5
+ checkpoint_name: resnet18
6
+ mix_choice: all_logits
7
+ data_types:
8
+ - image
9
+ train_transforms:
10
+ - resize_shorter_side
11
+ - center_crop
12
+ - trivial_augment
13
+ val_transforms:
14
+ - resize_shorter_side
15
+ - center_crop
16
+ image_norm: imagenet
17
+ image_size: null
18
+ image_chan_num: 3
19
+ use_learnable_image: false
20
+ max_image_num_per_column: 1
21
+ data:
22
+ image:
23
+ missing_value_strategy: zero
24
+ text:
25
+ normalize_text: false
26
+ categorical:
27
+ minimum_cat_count: 100
28
+ maximum_num_cat: 20
29
+ convert_to_text: false
30
+ convert_to_text_template: latex
31
+ numerical:
32
+ convert_to_text: false
33
+ scaler_with_mean: true
34
+ scaler_with_std: true
35
+ document:
36
+ missing_value_strategy: zero
37
+ label:
38
+ numerical_preprocessing: standardscaler
39
+ pos_label: null
40
+ column_features_pooling_mode: concat
41
+ mixup:
42
+ turn_on: false
43
+ mixup_alpha: 0.8
44
+ cutmix_alpha: 1.0
45
+ cutmix_minmax: null
46
+ prob: 1.0
47
+ switch_prob: 0.5
48
+ mode: batch
49
+ turn_off_epoch: 5
50
+ label_smoothing: 0.1
51
+ modality_dropout: 0
52
+ templates:
53
+ turn_on: false
54
+ num_templates: 30
55
+ template_length: 2048
56
+ preset_templates:
57
+ - super_glue
58
+ - rte
59
+ custom_templates: null
60
+ optim:
61
+ optim_type: adamw
62
+ lr: 0.0004
63
+ weight_decay: 0.001
64
+ lr_choice: layerwise_decay
65
+ lr_decay: 0.9
66
+ lr_schedule: cosine_decay
67
+ max_epochs: 20
68
+ max_steps: -1
69
+ warmup_steps: 0.1
70
+ end_lr: 0
71
+ lr_mult: 1
72
+ patience: 10
73
+ val_check_interval: 0.5
74
+ check_val_every_n_epoch: 1
75
+ skip_final_val: false
76
+ gradient_clip_val: 1
77
+ gradient_clip_algorithm: norm
78
+ track_grad_norm: -1
79
+ log_every_n_steps: 10
80
+ label_smoothing: 0
81
+ top_k: 3
82
+ top_k_average_method: greedy_soup
83
+ peft: null
84
+ lora:
85
+ module_filter: null
86
+ filter:
87
+ - query
88
+ - value
89
+ - ^q$
90
+ - ^v$
91
+ - ^k$
92
+ - ^o$
93
+ r: 8
94
+ alpha: 8
95
+ conv_lora_expert_num: 8
96
+ loss_func: auto
97
+ focal_loss:
98
+ alpha: null
99
+ gamma: 2.0
100
+ reduction: mean
101
+ mask2former_loss:
102
+ loss_cross_entropy_weight: 10.0
103
+ loss_mask_weight: 5.0
104
+ loss_dice_weight: 5.0
105
+ extra_trainable_params: []
106
+ cross_modal_align: null
107
+ cross_modal_align_weight: 0
108
+ automatic_optimization: true
109
+ lemda:
110
+ turn_on: false
111
+ arch_type: mlp_vae
112
+ z_dim: 8
113
+ num_layers: 6
114
+ kld_weight: 0.1
115
+ mse_weight: 0.1
116
+ adv_weight: 0.0001
117
+ consist_weight: 0.01
118
+ consist_threshold: 0.5
119
+ lr: 0.0001
120
+ optim_type: adamw
121
+ weight_decay: 1.0e-05
122
+ env:
123
+ num_gpus: 1
124
+ num_nodes: 1
125
+ batch_size: 128
126
+ per_gpu_batch_size: 8
127
+ inference_batch_size_ratio: 4
128
+ precision: 16-mixed
129
+ num_workers: 2
130
+ num_workers_inference: 2
131
+ accelerator: auto
132
+ fast_dev_run: false
133
+ deterministic: false
134
+ auto_select_gpus: true
135
+ strategy: auto
136
+ deepspeed_allgather_size: 1000000000.0
137
+ deepspeed_allreduce_size: 1000000000.0
138
+ compile:
139
+ turn_on: false
140
+ mode: default
141
+ dynamic: true
142
+ backend: inductor
data_processors.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bc0cc1c584d582f5e79c7dc2d1ebd08c875cf4e9afbde68fb2987c877b6a021c
3
+ size 1517
df_preprocessor.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:574cc842d45e8e736935ccd2a1b742b47239b2ed211874274d18905afc16d38b
3
+ size 13375
eval_metric.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b0024ce25367aac2c6595f96f9b18415e8483aa9fb4771dff28ea934a6ce1bca
3
+ size 220
events.out.tfevents.1758242592.8d24268efa0f.10176.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:13d57c0d658161eb9f43b7403b6f07e3886c800fa1228accd2e15484438dea05
3
+ size 8805
hparams.yaml ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ optim_type: adamw
2
+ lr_choice: layerwise_decay
3
+ lr_schedule: cosine_decay
4
+ lr: 0.0004
5
+ lr_decay: 0.9
6
+ end_lr: 0
7
+ lr_mult: 1
8
+ weight_decay: 0.001
9
+ warmup_steps: 0.1
10
+ validation_metric_name: accuracy
11
+ peft: null
12
+ mixup_off_epoch: 5
13
+ skip_final_val: false
14
+ track_grad_norm: -1
15
+ cross_modal_align: null
16
+ cross_modal_align_weight: 0
17
+ automatic_optimization: true
18
+ accumulate_grad_batches: 16
19
+ gradient_clip_val: 1
20
+ gradient_clip_algorithm: norm
21
+ use_aug_optim: false
22
+ aug_lr: 0.0001
23
+ aug_weight_decay: 1.0e-05
24
+ aug_optim_type: adamw
model.ckpt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:883cd2981aabafe758f632a4d5e6ff0f7397bd3f4196a08692a3377fda5428b3
3
+ size 44779915