Hugging Face's logo

microsoft
/
SNRAware

English
Model card Files
xet
 Community
SNRAware
File size: 4,582 Bytes
5d5e025
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
backbone:
  num_of_channels: 64
  block_str:
  - T1L1G1T1L1G1T1L1G1T1L1G1
  - T1L1G1T1L1G1T1L1G1T1L1G1
  block:
    cell:
      attention_type: conv
      mixer_type: conv
      window_size:
      - 8
      - 8
      - 16
      patch_size:
      - 4
      - 4
      - 2
      window_sizing_method: mixed
      n_head: 64
      scale_ratio_in_mixer: 4.0
      normalize_Q_K: true
      cosine_att: true
      att_with_relative_position_bias: true
      att_dropout_p: 0.0
      dropout_p: 0.1
      att_with_output_proj: true
      norm_mode: layer
      activation_func: prelu
      upsample_method: linear
      with_timer: false
      temporal:
        _target_: ifm.model.config.TemporalAttentionConfig
        stride_qk:
        - 1
        - 1
        - 1
      spatial_local:
        _target_: ifm.model.config.SpatialLocalConfig
      spatial_global:
        _target_: ifm.model.config.SpatialGlobalConfig
        shuffle_in_window: false
      convoluation:
        _target_: ifm.model.config.ConvolutionConfig
        conv_type: conv3d
      spatial_local_3d:
        _target_: ifm.model.config.SpatialLocal3DConfig
      spatial_global_3d:
        _target_: ifm.model.config.SpatialGlobal3DConfig
        shuffle_in_window: false
      spatial_vit:
        _target_: ifm.model.config.SpatialViTConfig
      vit_3d:
        _target_: ifm.model.config.ViT3DConfig
      swin_3d:
        _target_: ifm.model.config.Swin3DConfig
    _target_: ifm.model.config.BlockConfig
    cell_type: sequential
    block_dense_connection: false
  _target_: ifm.model.config.HRNetConfig
  name: HRnet
  num_resolution_levels: 2
optim:
  lr: 1.0e-05
  weight_decay: 0.0
  name: sophia
  beta1: 0.9
  beta2: 0.999
  eps: 1.0e-08
  rho: 0.01
scheduler:
  name: OneCycleLR
  pct_start: 0.3
  anneal_strategy: cos
  div_factor: 25
dataset:
  _target_: ifm.mri.denoising.data.MRIDenoisingDataset
  cutout_shape:
  - 64
  - 64
  - 16
  repetition: 1
  min_noise_level: 0.1
  max_noise_level: 128.0
  kspace_filter_sigma:
  - 0.8
  - 1.0
  - 1.5
  - 2.0
  - 2.25
  kspace_filter_sigma_T:
  - 0.25
  - 0.5
  - 0.65
  - 0.85
  - 1.0
  - 1.5
  - 2.0
  - 2.25
  prob_apply_filter_T: 0.2
  pf_filter_ratio:
  - 1.0
  - 0.875
  - 0.75
  - 0.625
  phase_resolution_ratio:
  - 1.0
  - 0.75
  - 0.65
  - 0.55
  readout_resolution_ratio:
  - 1.0
  - 0.75
  - 0.65
  - 0.55
  only_white_noise: false
  ignore_gmap: false
  add_noise: true
  add_salt_pepper: true
  salt_pepper_amount: 0.4
  salt_pepper_prob: 0.4
  add_poisson: true
  poisson_prob: 0.4
  shuffle_along_3rd_dim: true
  shuffle_along_3rd_dim_prob: 0.1
  matrix_size_adjust_ratio:
  - 0.5
  - 0.75
  - 1.0
  - 1.25
  - 1.5
  matrix_size_adjust_prob: 0.5
  resolution_reduction_prob: 0.0
  single_frame_mode: true
  single_frame_mode_prob: 0.1
  dicom_mode: false
trainer:
  _target_: lightning.Trainer
  max_epochs: 320
  precision: 32
  devices: 8
  accelerator: gpu
  strategy: ddp
  num_nodes: 2
  gradient_clip_val: 1.0
  accumulate_grad_batches: 1
  log_every_n_steps: 200
  enable_progress_bar: true
  enable_model_summary: true
  enable_checkpointing: true
  check_val_every_n_epoch: 16
  val_check_interval: 1.0
  fast_dev_run: false
  overfit_batches: 0.0
  limit_train_batches: 1.0
  limit_val_batches: 1.0
  limit_test_batches: 1.0
  deterministic: false
  benchmark: true
  profiler: null
logging:
  project: ifm-mri-denoising
  run_name: ifm-mri-denoising-large-no-epoch
  output_dir: /data/ifm-mri-denoising/ifm-mri-denoising_20251004_025456/ifm-mri-denoising-large-no-epoch
  wandb_entity: biomed-signal-processing
  wandb_dir: /data/ifm-mri-denoising/ifm-mri-denoising_20251004_025456/ifm-mri-denoising-large-no-epoch/wandb
  use_wandb: true
  save_ckpt_every_n_epochs: 16
  save_batches_to_output_dir: false
  log_train_batches: 16
  log_val_batches: 16
  log_test_batches: 32
tyger:
  mode: false
  dop: 8
step_mode:
  total_steps: 0
  steps_to_report: 200
  steps_to_timing: 1000
  steps_to_validate: 10000
  steps_to_save_ckpts: 20000
train_data_dir: /data/imaging-fm-projects/mri/denoising/data/whole/tra
test_data_dir: /data/imaging-fm-projects/mri/denoising/data/whole/test
seed: 3782770815
max_epochs: 320
batch_size: 2
num_workers: 8
prefetch_factor: 8
loss:
- mse
- perpendicular
- perceptual
- charbonnier
- gaussian3d
loss_weights:
- 1.0
- 1.0
- 1.0
- 1.0
- 1.0
val_data_portion: 0.02
overlap_for_inference:
- 16
- 16
- 8
device_mesh:
- 8
- 2
date: rank_0_544387_02-57-14-20251004
rank: 0
global_rank: 0
rank_str: "\e[34m\e[47mrank 0, global_rank 0 \e[0m"
device: cuda:0
yaml_file: null
log_file_name: null
ddp: true