-
Notifications
You must be signed in to change notification settings - Fork 95
/
Copy pathscDINO_full_pipeline.yaml
69 lines (67 loc) · 1.5 KB
/
scDINO_full_pipeline.yaml
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
meta:
name_of_run: "scDino_pcluster_full_0"
output_dir: "/fsx/outputdir"
dataset_dir: "/fsx/data"
folder_depth_for_labels: 0
train_datasetsplit_fraction: 0.8
seed: 40
images_are_RGB: False
channel_dict: "aTub,BF,DAPI,Oct4,PE"
selected_channel_combination_per_run: 0,1,2,3,4
center_crop: 0
train_scDINO:
dino_vit_name: "mySCDINOmodel"
fraction_for_mean_std_calc: 0.2
epochs: 3
saveckp_freq: 1
num_gpus: 4
nnodes: 2
batch_size_per_gpu: 30
num_workers: 16
dist_url: "env://"
hyperparameters:
arch: "vit_small"
patch_size: 16
norm_last_layer: True
momentum_teacher: 0.996
use_bn_in_head: False
warmup_teacher_temp: 0.04
techer_temp: 0.04
warmup_teacher_temp_epochs: 0
use_fp16: True
weight_decay: 0.04
weight_decay_end: 0.4
clip_grad: 3
freeze_last_layer: 1
lr: 0.0005
warump_epochs: 1
min_lr: 0.000001
optimizer: "adamw"
drop_path_rate: 0.1
local_crops_number: 8
downstream_analyses:
compute_cls_features:
normalize: True
num_gpus: 4
nnodes: 2
batch_size_per_gpu: 24
num_workers: 16
checkpoint_key: teacher
resize: True
resize_length: 224
attention_visualisation:
num_images_per_class: 1
resize_attention_image: True
image_size: 480
kNN:
global:
n_neighbors: [5, 10, 50, 100, 200, 500]
temperature: 0.1
umap_eval:
n_neighbors: 30
min_dist: 0.4
n_components: 2
metric: euclidean
spread: 1.1
epochs: 100
topometry_plots: False