sync configs/kermany/fm_stage2/DiTDH-XL_VisionFM.yaml

configs/kermany/fm_stage2/DiTDH-XL_VisionFM.yaml

@@ -0,0 +1,90 @@
+stage_1:
+  target: stage1.RAE
+  params:
+    encoder_cls: VisionFM
+    encoder_config_path: /mnt/tidal-alsh-share2/dataset/qinshengqian/research/c3/Data/Classification/OCT_8/Baseline/RAE-main/configs/encoders/visionfm
+    encoder_input_size: 256
+    encoder_params:
+      weight_path: /mnt/tidal-alsh-share2/dataset/qinshengqian/research/c3/Data/Classification/OCT_8/Baseline/VisionFM/VFM_OCT_weights.pth
+      encoder_input_size: 256
+    decoder_config_path: /mnt/tidal-alsh-share2/dataset/qinshengqian/research/c3/Data/Classification/OCT_8/Baseline/RAE-main/configs/decoder/ViTXL
+    decoder_patch_size: 16
+    pretrained_decoder_path: /mnt/tidal-alsh-share2/dataset/qinshengqian/research/c3/Data/Classification/OCT_8/Baseline/RAE-main/results_kermany/stage1/stage1_VisionFM/checkpoints/decoder_ema.pt
+    noise_tau: 0.0
+    reshape_to_2d: true
+    normalization_stat_path: /mnt/tidal-alsh-share2/dataset/qinshengqian/research/c3/Data/Classification/OCT_8/Baseline/RAE-main/results_kermany/stats/VisionFM/normalization_stats.pt
+stage_2:
+  target: stage2.models.DDT.DiTwDDTHead
+  params:
+    input_size: 16
+    patch_size: 1
+    in_channels: 768
+    hidden_size:
+    - 1152
+    - 2048
+    depth:
+    - 28
+    - 2
+    num_heads:
+    - 16
+    - 16
+    mlp_ratio: 4.0
+    class_dropout_prob: 0.1
+    num_classes: 4
+    use_qknorm: false
+    use_swiglu: true
+    use_rope: true
+    use_rmsnorm: true
+    wo_shift: false
+    use_pos_embed: true
+transport:
+  params:
+    path_type: Linear
+    prediction: velocity
+    loss_weight: null
+    time_dist_type: logit-normal_0_1
+sampler:
+  mode: ODE
+  params:
+    sampling_method: euler
+    num_steps: 50
+    atol: 1.0e-06
+    rtol: 0.001
+    reverse: false
+guidance:
+  method: cfg
+  scale: 1.0
+  t_min: 0.0
+  t_max: 1.0
+misc:
+  latent_size:
+  - 768
+  - 16
+  - 16
+  num_classes: 4
+  time_dist_shift_dim: 196608
+  time_dist_shift_base: 4096
+training:
+  epochs: 50
+  global_batch_size: 64
+  grad_accum_steps: 1
+  ema_decay: 0.9995
+  num_workers: 8
+  log_interval: 100
+  checkpoint_interval: 20
+  sample_every: 5000
+  clip_grad: 1.0
+  global_seed: 42
+  optimizer:
+    lr: 0.0002
+    betas:
+    - 0.9
+    - 0.95
+    weight_decay: 0.0
+  scheduler:
+    type: linear
+    warmup_epochs: 10
+    decay_end_epoch: 50
+    base_lr: 0.0002
+    final_lr: 2.0e-05
+    warmup_from_zero: false