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
Debug Loss: target_masks shape before resize: torch.Size([1, 8, 8])
Epoch 50/50 | Batch 10820/10821 | Loss: 0.0003
loss_ce: 0.0003
loss_mask: 0.0000
loss_dice: 0.0000
Debug Dataset: no annotations, creating placeholder with mask_size: (8, 8)
Debug Dataset: final masks shape: torch.Size([1, 8, 8])
Debug Dataset: final labels shape: torch.Size([1])
Debug Train: imgs shape: torch.Size([1, 3, 128, 128])
Debug Train: target 0 masks shape: torch.Size([1, 8, 8])
Debug Train: target 0 labels shape: torch.Size([1])
Debug: VAE latents shape: torch.Size([1, 4, 16, 16])
Debug: ODISEModel input shape: torch.Size([1, 4, 16, 16])
Debug: UNetEncoder input latent shape: torch.Size([1, 4, 16, 16])
Debug: After conv_in: torch.Size([1, 320, 16, 16])
Debug: After down_block 0: torch.Size([1, 320, 8, 8])
Debug: After down_block 1: torch.Size([1, 640, 4, 4])
Debug: After down_block 2: torch.Size([1, 1280, 2, 2])
Debug: After down_block 3: torch.Size([1, 1280, 2, 2])
Debug: UNet encoded features shape: torch.Size([1, 1280, 2, 2])
Debug: PixelDecoder input shape: torch.Size([1, 1280, 2, 2])
Debug: After input_proj: torch.Size([1, 256, 2, 2])
Debug: After upsample 0: torch.Size([1, 256, 4, 4])
Debug: After upsample 1: torch.Size([1, 256, 8, 8])
Debug: Final mask_features shape: torch.Size([1, 256, 8, 8])
Debug: Final outputs_mask shape: torch.Size([1, 10, 8, 8])
Debug Train: outputs pred_logits shape: torch.Size([1, 10, 2])
Debug Train: outputs pred_masks shape: torch.Size([1, 10, 8, 8])
Debug Matcher: pred_logits shape: torch.Size([1, 10, 2])
Debug Matcher: pred_masks shape: torch.Size([1, 10, 8, 8])
Debug Matcher: out_prob shape: torch.Size([10, 2])
Debug Matcher: out_mask shape: torch.Size([10, 8, 8])
Debug Matcher: tgt_ids shape: torch.Size([1])
Debug Matcher: tgt_mask shape: torch.Size([1, 8, 8])
Debug Matcher: out_mask_flat shape: torch.Size([10, 64])
Debug Matcher: tgt_mask_flat shape: torch.Size([1, 64])
Debug Loss: src_masks shape: torch.Size([1, 8, 8])
Debug Loss: target_masks shape before resize: torch.Size([1, 8, 8])
Epoch 50/50 完成 | Avg Loss: 0.0525
訓練完成!
(base) jovyan@unzip-workspace-0-2:/mnt/nfs/nina/nina/segg$ ls
segg.py segmentation_output
(base) jovyan@unzip-workspace-0-2:/mnt/nfs/nina/nina/segg$ cd segmentation_output
(base) jovyan@unzip-workspace-0-2:/mnt/nfs/nina/nina/segg/segmentation_output$ ls
best_odise_model.pth final_odise_model.pth
(base) jovyan@unzip-workspace-0-2:/mnt/nfs/nina/nina/segg/segmentation_output$