Skip to content

Refactor/change checkpoint loading mechanism - #20

Open
haovo3326 wants to merge 54 commits into
lyp-deeplearning:mainfrom
haovo3326:refactor/change-checkpoint-loading-mechanism
Open

haovo3326 wants to merge 54 commits into
lyp-deeplearning:mainfrom
haovo3326:refactor/change-checkpoint-loading-mechanism

Conversation

@haovo3326

Copy link
Copy Markdown

No description provided.

feat: implement moge_wrapper for depth/normal injection
feat: introduce 1x1 convolution and residual into 3D-GFL
Removed 'third_repos/' from the .gitignore file.
refactor: change file paths and loading mechanism to fit Kaggle
refactor: change file paths in ALIKE and Depth-Anything-v2 to adapt t…
refactor: update ALIKE and DepthAnythingV2 system path to fit Kaggle
refactor: disable setproctitle package in train.py
refactor: remove duplicate training log printing and pretrained model…
refactor: adopted M1 architectural design in featurebooster.py
refactor: adopted M1 architectural config
Comment out the loading of the pretrained model.
…tention

feat: intergrate multi-head attention in 3D GFL
Comment out the loading of the pretrained model in train.py.
…tention

hotfix: fix d_model naming error in loss.py
feat: saving optimizer state dict for discontinuous training
feat: skip batch if data loading is failed
…oading

feat: load the latest checkpoint in ./weights for continuous training
…cture

fix: change model loading to adapt to new architecture with different…
feat: add host ID input for fast kaggle account switching
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.

1 participant