Added support for lightning and hydra

.gitignore

@@ -1,2 +1,169 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#pdm.lock
+#   pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
+#   in version control.
+#   https://pdm.fming.dev/#use-with-ide
+.pdm.toml
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  and can be added to the global gitignore or merged into this file.  For a more nuclear
+#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
+#.idea/
+
+
+.vscode/
+checkpoints/
+wandb/
+outputs/
+datasets/*
 /logs/
 /checkpoints/

src/configs/computer/1-gpu.yaml

@@ -0,0 +1,7 @@
+devices: 1
+progress_bar_refresh_rate: 2
+num_workers: 8
+accelerator: gpu
+precision: bf16-mixed
+strategy: auto
+num_nodes: 1

src/configs/computer/2-gpu.yaml

@@ -0,0 +1,7 @@
+devices: 2
+progress_bar_refresh_rate: 2
+num_workers: 8
+accelerator: gpu
+precision: bf16-mixed
+strategy: ddp
+num_nodes: 1

src/configs/computer/4-gpu.yaml

@@ -0,0 +1,7 @@
+devices: 4
+progress_bar_refresh_rate: 2
+num_workers: 8
+accelerator: gpu
+precision: bf16-mixed
+strategy: ddp
+num_nodes: 1

src/configs/data/additional_train_transforms/randaugment.yaml

@@ -0,0 +1,11 @@
+instance:
+  - _target_: torchvision.transforms.v2.RandomApply
+    transforms:
+      - _target_: torchvision.transforms.v2.RandAugment
+        magnitude: ${data.additional_train_transforms.randaugment_magnitude}
+    p: ${data.additional_train_transforms.randaugment_p}
+
+
+
+randaugment_magnitude: 6
+randaugment_p: 0.1

src/configs/data/imagenet.yaml

@@ -0,0 +1,24 @@
+defaults:
+  - additional_train_transforms: randaugment
+  - _self_
+
+full_batch_size: 128
+size: 256
+num_classes: 1000
+name: Imagenet-1K
+
+dataset_builder:
+  _partial_: true
+  _target_: utils.data.build_imagenet
+  data_dir: ${data_dir}/imagenet
+  num_classes: ${data.num_classes}
+  size: ${data.size}
+  additional_transforms: ${data.additional_train_transforms.instance}
+
+datamodule:
+  _target_: utils.datamodule.ImageDataModule
+  dataset_builder: ${data.dataset_builder}
+  full_batch_size: ${data.full_batch_size}
+  num_workers: ${computer.num_workers}
+  num_nodes: ${computer.num_nodes}
+  num_devices: ${computer.devices}

src/configs/logger/wandb.yaml

@@ -0,0 +1,6 @@
+_target_: pytorch_lightning.loggers.WandbLogger
+save_dir: ${root_dir}/wandb
+name: ${experiment_name}
+project: Homm
+log_model: False
+offline: False

src/configs/model/homm.yaml

@@ -0,0 +1,24 @@
+defaults:
+  - train_batch_preprocess: mixup_and_cutmix  
+  - network: homm
+  - loss: CE
+  - optimizer: adamw
+  - lr_scheduler: warmup_cosine_decay
+  - train_metric: accuracy
+  - val_metric: accuracy
+  - test_metric: accuracy
+
+name: HoMM
+instance:
+  _target_: model.classification.ClassificationModule
+  model: ${model.network}
+  loss: ${model.loss}
+  optimizer_cfg: ${model.optimizer}
+  lr_scheduler_builder: ${model.lr_scheduler}
+  train_batch_preprocess: ${model.train_batch_preprocess}
+  train_metrics: ${model.train_metric}
+  val_metrics: ${model.val_metric}
+  test_metrics: ${model.test_metric}
+
+
+lsuv_normalize: False

src/configs/model/loss/CE.yaml

@@ -0,0 +1,2 @@
+_target_: model.losses.CE.CrossEntropyLossModule
+sum_label_dim: True

src/configs/model/lr_scheduler/warmup.yaml

@@ -0,0 +1,4 @@
+_partial_: true
+_target_: utils.lr_scheduler.WarmupLR
+warmup_steps: 10000
+

src/configs/model/lr_scheduler/warmup_cosine_decay.yaml

@@ -0,0 +1,5 @@
+_partial_: true
+_target_: utils.lr_scheduler.WarmupCosineDecayLR
+warmup_steps: 10000
+total_steps: ${trainer.max_steps}
+

src/configs/model/network/homm.yaml

@@ -0,0 +1,12 @@
+_target_: model.network.vision.HoMVision
+nb_classes: ${data.num_classes}
+dim: 256
+im_size: 256
+kernel_size: 32
+nb_layers: 4
+order: 2
+order_expand: 4
+ffw_expand: 4
+dropout: 0.0
+pooling: cls
+in_conv: true

src/configs/model/optimizer/adam.yaml

@@ -0,0 +1,8 @@
+optim:
+  _partial_: true
+  _target_: torch.optim.Adam
+  lr: 1e-3
+  betas: [0.9, 0.999]
+  weight_decay: 0.01
+
+exclude_ln_and_biases_from_weight_decay: False

src/configs/model/optimizer/adamw.yaml

@@ -0,0 +1,8 @@
+optim:
+  _partial_: true 
+  _target_: torch.optim.AdamW
+  lr: 1e-3
+  betas: [0.9, 0.999]
+  weight_decay: 0.01
+
+exclude_ln_and_biases_from_weight_decay: False

src/configs/model/optimizer/sgd.yaml

@@ -0,0 +1,7 @@
+optim:
+  _partial_: true
+  _target_: torch.optim.SGD
+  lr: 1e-3
+  weight_decay: 0.01
+
+exclude_ln_and_biases_from_weight_decay: False

src/configs/model/test_metric/accuracy.yaml

@@ -0,0 +1,2 @@
+_target_: utils.metrics.ClassificationMetrics
+num_classes: ${data.num_classes}

src/configs/model/train_batch_preprocess/cutmix.yaml

@@ -0,0 +1,4 @@
+_target_: utils.mixup.CutMix
+apply_transform_prob: 1.0
+alpha: 1.0
+num_classes: 1000

src/configs/model/train_batch_preprocess/mixup.yaml

@@ -0,0 +1,4 @@
+_target_: utils.mixup.MixUp
+apply_transform_prob: 1.0
+alpha: 0.1
+num_classes: 1000

src/configs/model/train_batch_preprocess/mixup_and_cutmix.yaml

@@ -0,0 +1,6 @@
+_target_: utils.mixup.CutMixUp
+apply_transform_prob: 1.0
+mixup_prob: 0.5
+alpha_mixup: 0.1
+alpha_cutmix: 1.0
+num_classes: 1000

src/configs/model/train_metric/accuracy.yaml

@@ -0,0 +1,2 @@
+_target_: utils.metrics.ClassificationMetrics
+num_classes: ${data.num_classes}

src/configs/model/val_metric/accuracy.yaml

@@ -0,0 +1,2 @@
+_target_: utils.metrics.ClassificationMetrics
+num_classes: ${data.num_classes}

src/configs/train.yaml

@@ -0,0 +1,42 @@
+defaults:
+  - model: homm
+  - computer: 1-gpu
+  - data: imagenet
+  - logger: wandb
+  - _self_
+
+trainer:
+  _target_: pytorch_lightning.Trainer
+  max_steps: 300000
+  deterministic: True
+  devices: ${computer.devices}
+  accelerator: ${computer.accelerator}
+  strategy: ${computer.strategy}
+  log_every_n_steps: 1
+  num_nodes: ${computer.num_nodes}
+  precision: ${computer.precision}
+  gradient_clip_val: 1.0
+
+checkpoints:
+  _target_: pytorch_lightning.callbacks.ModelCheckpoint
+  dirpath: ${checkpoint_dir}/${experiment_name}
+  save_last: True
+  monitor: val/loss
+  mode: min
+
+progress_bar:
+  _target_: pytorch_lightning.callbacks.TQDMProgressBar
+  refresh_rate: ${computer.progress_bar_refresh_rate}
+
+seed: 3407
+data_dir: ${root_dir}/datasets
+root_dir:  ${hydra:runtime.cwd}
+checkpoint_dir: ${root_dir}/checkpoints
+experiment_name_suffix: base
+experiment_name: ${data.name}_${model.name}_${experiment_name_suffix}
+
+hydra:
+  run:
+    dir: outputs/${hydra.job.name}/${now:%Y-%m-%d_%H-%M-%S}/${experiment_name}
+  job:
+    chdir: true