Skip to content

fix channel mismatch in new_model.py when i==1#47

Open
JingweiZhang12 wants to merge 15 commits intoNoamRosenberg:masterfrom
JingweiZhang12:master
Open

fix channel mismatch in new_model.py when i==1#47
JingweiZhang12 wants to merge 15 commits intoNoamRosenberg:masterfrom
JingweiZhang12:master

Conversation

@JingweiZhang12
Copy link

  1. discard useless parameter 'criterion' in Class newModel
  2. fix channel mismatch in new_model.py when i==1
  3. change the default cityscapes epoch to 90
  4. remove args.gpu_ids, replace the 'len(agrs.gpu_ids)' by 'torch.cuda.device_count()'
  5. some parameters changes considering practical training, not matter

JingweiZhang12 and others added 14 commits August 31, 2019 17:02
1. fix the error: only 11 layers are passed
2.Set cudnn.bechmark for slightly speeding up.
3. Remove unused argument.
4. epoch 0~19 train weights only and epoch 20~39 train both weights and archs.
fix channel mismatch by add self._filter_multiplier * filter_param_dict[prev_level] in new_model
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants