├── README.md ├── precision-benchmarks ├── torchvision-vit16-small │ ├── requirements.txt │ ├── README.md │ ├── 07_bf16-full-with-trainer.py │ ├── local_utilities.py │ ├── trainer_utilities.py │ ├── 01_pytorch-fp32.py │ ├── 02_pytorch-fabric-fp32.py │ ├── 03_fp16-mixed.py │ ├── 04_bf16-mixed.py │ ├── 05_fp16-full.py │ └── 06_bf16-full.py └── torchvision-vit16-large │ ├── README.md │ ├── 07_bf16-full-with-trainer.py │ ├── 04_trainer.py │ ├── local_utilities.py │ ├── trainer_utilities.py │ ├── 01_pytorch-fp32.py │ ├── 02_pytorch-fabric-fp32.py │ ├── 03_fp16-mixed.py │ ├── 05_fp16-full.py │ ├── 06_bf16-full.py │ └── 04_bf16-mixed.py ├── .gitignore └── LICENSE /README.md: -------------------------------------------------------------------------------- 1 | # ViT-finetuning-scripts 2 | Vision transformer finetuning scripts 3 | -------------------------------------------------------------------------------- /precision-benchmarks/torchvision-vit16-small/requirements.txt: -------------------------------------------------------------------------------- 1 | numpy >= 1.24.3 2 | scipy >= 1.10.1 3 | pandas >= 2.0.2 4 | watermark >= 2.4.2 5 | torch >= 2.1.0 6 | torchvision >= 0.15.2 7 | torchmetrics >= 0.11.4 8 | transformers >= 4.30.2 9 | lightning >= 2.1.0 10 | deepspeed >= 0.9.4 -------------------------------------------------------------------------------- /precision-benchmarks/torchvision-vit16-small/README.md: -------------------------------------------------------------------------------- 1 | | Script | Model | Matmul precision | Runtime | Memory | Train accuracy | Test accuracy | 2 | | ------------------------------------------------------------ | -------- | ---------------- | -------- | ------- | -------------- | ------------- | 3 | | [01_pytorch-fp32.py](http://01_pytorch-fp32.py) | vit_b_16 | medium | 7.71 min | 3.71 GB | 97.96% | 95.27% | 4 | | [02_pytorch-fabric-fp32.py](http://02_pytorch-fabric-fp32.py) | vit_b_16 | medium | 7.53 min | 3.71 GB | 97.87% | 95.54% | 5 | | [03_fp16-mixed.py](http://03_fp16-mixed.py) | vit_b_16 | medium | 9.38 min | 3.03 GB | 97.94% | 96.09% | 6 | | [04_bf16-mixed.py](http://04_bf16-mixed.py/) | vit_b_16 | medium | 8.50 min | 3.03 GB | 97.86% | 95.16% | 7 | | [05_fp16-full.py](http://05_fp16-full.py) | vit_b_16 | medium | 7.22 min | 1.94 GB | 10.01% | 10.00% | 8 | | [06_bf16-full.py](http://06_bf16-full.py) | vit_b_16 | medium | 7.00 min | 1.95 GB | 99.69% | 97.52% | 9 | 10 | | PyTorch: 2.1.0+cu121 | 11 | | -------------------- | 12 | | Lightning: 2.1.0 | -------------------------------------------------------------------------------- /precision-benchmarks/torchvision-vit16-large/README.md: -------------------------------------------------------------------------------- 1 | | Script | Model | Matmul precision | Runtime | Memory | Train accuracy | Test accuracy | 2 | | ------------------------------------------------------------ | -------- | ---------------- | --------- | -------- | -------------- | ------------- | 3 | | [01_pytorch-fp32.py](http://01_pytorch-fp32.py) | vit_l_16 | medium | 16.88 min | 16.70 GB | 98.40% | 94.06% | 4 | | [02_pytorch-fabric-fp32.py](http://02_pytorch-fabric-fp32.py) | vit_l_16 | medium | 17.03 min | 16.70 GB | 98.49% | 96.17% | 5 | | [03_fp16-mixed.py](http://03_fp16-mixed.py) | vit_l_16 | medium | 11.63 min | 12.30 GB | 98.47% | 94.79% | 6 | | [04_bf16-mixed.py](http://04_bf16-mixed.py/) | vit_l_16 | medium | 11.31 min | 12.24 GB | 98.46% | 95.62% | 7 | | [05_fp16-full.py](http://05_fp16-full.py) | vit_l_16 | medium | 9.19 min | 8.43 GB | 10.02% | 10.00% | 8 | | [06_bf16-full.py](http://06_bf16-full.py) | vit_l_16 | medium | 9.37 min | 8.43 GB | 99.92% | 97.86% | 9 | 10 | | PyTorch: 2.1.0+cu121 | 11 | | -------------------- | 12 | | Lightning: 2.1.0 | -------------------------------------------------------------------------------- /precision-benchmarks/torchvision-vit16-large/07_bf16-full-with-trainer.py: -------------------------------------------------------------------------------- 1 | import lightning as L 2 | import torch 3 | from torchvision.models import vit_l_16 4 | from torchvision.models import ViT_L_16_Weights 5 | from torchvision import transforms 6 | 7 | from shared_utilities import get_dataloaders_cifar10 8 | from trainer_utilities import LightningModel 9 | 10 | #################### 11 | # Initialize Model 12 | #################### 13 | 14 | L.seed_everything(123) 15 | pytorch_model = vit_l_16(weights=ViT_L_16_Weights.IMAGENET1K_V1) 16 | 17 | # replace output layer 18 | pytorch_model.heads.head = torch.nn.Linear(in_features=1024, out_features=10) 19 | 20 | 21 | #################### 22 | # Load Dataset 23 | #################### 24 | 25 | train_transforms = transforms.Compose([transforms.Resize((224, 224)), 26 | # transforms.RandomCrop((224, 224)), 27 | transforms.ToTensor()]) 28 | 29 | test_transforms = transforms.Compose([transforms.Resize((224, 224)), 30 | # transforms.CenterCrop((224, 224)), 31 | transforms.ToTensor()]) 32 | train_loader, val_loader, test_loader = get_dataloaders_cifar10( 33 | batch_size=32, 34 | num_workers=4, 35 | train_transforms=train_transforms, 36 | test_transforms=test_transforms, 37 | validation_fraction=0.1, 38 | download=True 39 | ) 40 | 41 | #################### 42 | # Train Model 43 | #################### 44 | 45 | lightning_model = LightningModel(model=pytorch_model, learning_rate=5e-5) 46 | 47 | trainer = L.Trainer( 48 | max_epochs=3, 49 | accelerator="gpu", 50 | devices=1, 51 | precision="bf16-true", 52 | ) 53 | 54 | trainer.fit(model=lightning_model, train_dataloaders=train_loader, val_dataloaders=val_loader) 55 | trainer.test(model=lightning_model, dataloaders=test_loader) -------------------------------------------------------------------------------- /precision-benchmarks/torchvision-vit16-small/07_bf16-full-with-trainer.py: -------------------------------------------------------------------------------- 1 | import lightning as L 2 | import torch 3 | from torchvision.models import vit_b_16 4 | from torchvision.models import ViT_B_16_Weights 5 | from torchvision import transforms 6 | 7 | from shared_utilities import get_dataloaders_cifar10 8 | from trainer_utilities import LightningModel 9 | 10 | #################### 11 | # Initialize Model 12 | #################### 13 | 14 | L.seed_everything(123) 15 | pytorch_model = vit_b_16(weights=ViT_B_16_Weights.IMAGENET1K_V1) 16 | 17 | # replace output layer 18 | pytorch_model.heads.head = torch.nn.Linear(in_features=768, out_features=10) 19 | 20 | 21 | #################### 22 | # Load Dataset 23 | #################### 24 | 25 | train_transforms = transforms.Compose([transforms.Resize((224, 224)), 26 | # transforms.RandomCrop((224, 224)), 27 | transforms.ToTensor()]) 28 | 29 | test_transforms = transforms.Compose([transforms.Resize((224, 224)), 30 | # transforms.CenterCrop((224, 224)), 31 | transforms.ToTensor()]) 32 | train_loader, val_loader, test_loader = get_dataloaders_cifar10( 33 | batch_size=32, 34 | num_workers=4, 35 | train_transforms=train_transforms, 36 | test_transforms=test_transforms, 37 | validation_fraction=0.1, 38 | download=True 39 | ) 40 | 41 | #################### 42 | # Train Model 43 | #################### 44 | 45 | lightning_model = LightningModel(model=pytorch_model, learning_rate=5e-5) 46 | 47 | trainer = L.Trainer( 48 | max_epochs=3, 49 | accelerator="gpu", 50 | devices=1, 51 | precision="bf16-true", 52 | ) 53 | 54 | trainer.fit(model=lightning_model, train_dataloaders=train_loader, val_dataloaders=val_loader) 55 | trainer.test(model=lightning_model, dataloaders=test_loader) -------------------------------------------------------------------------------- /precision-benchmarks/torchvision-vit16-large/04_trainer.py: -------------------------------------------------------------------------------- 1 | import lightning as L 2 | import torch 3 | from torchvision.models import vit_l_16 4 | from torchvision.models import ViT_L_16_Weights 5 | from torchvision import transforms 6 | 7 | from shared_utilities import get_dataloaders_cifar10 8 | from trainer_utilities import LightningModel 9 | 10 | #################### 11 | # Initialize Model 12 | #################### 13 | 14 | L.seed_everything(123) 15 | pytorch_model = vit_l_16(weights=ViT_L_16_Weights.IMAGENET1K_V1) 16 | 17 | # replace output layer 18 | pytorch_model.heads.head = torch.nn.Linear(in_features=1024, out_features=10) 19 | 20 | 21 | #################### 22 | # Load Dataset 23 | #################### 24 | 25 | train_transforms = transforms.Compose([transforms.Resize((224, 224)), 26 | # transforms.RandomCrop((224, 224)), 27 | transforms.ToTensor()]) 28 | 29 | test_transforms = transforms.Compose([transforms.Resize((224, 224)), 30 | # transforms.CenterCrop((224, 224)), 31 | transforms.ToTensor()]) 32 | train_loader, val_loader, test_loader = get_dataloaders_cifar10( 33 | batch_size=32, 34 | num_workers=4, 35 | train_transforms=train_transforms, 36 | test_transforms=test_transforms, 37 | validation_fraction=0.1, 38 | download=True 39 | ) 40 | 41 | #################### 42 | # Train Model 43 | #################### 44 | 45 | lightning_model = LightningModel(model=pytorch_model, learning_rate=5e-5) 46 | 47 | trainer = L.Trainer( 48 | max_epochs=3, 49 | accelerator="gpu", 50 | devices=1, 51 | precision="bf16-true", 52 | deterministic=True, 53 | ) 54 | 55 | trainer.fit(model=lightning_model, train_dataloaders=train_loader, val_dataloaders=val_loader) 56 | trainer.test(model=lightning_model, dataloaders=test_loader) -------------------------------------------------------------------------------- /precision-benchmarks/torchvision-vit16-large/local_utilities.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from torch.utils.data import sampler 3 | from torchvision import datasets 4 | from torch.utils.data import DataLoader 5 | from torch.utils.data import SubsetRandomSampler 6 | from torchvision import transforms 7 | 8 | 9 | def get_dataloaders_cifar10(batch_size, num_workers=0, 10 | validation_fraction=None, 11 | train_transforms=None, 12 | test_transforms=None, 13 | download=True): 14 | 15 | if train_transforms is None: 16 | train_transforms = transforms.ToTensor() 17 | 18 | if test_transforms is None: 19 | test_transforms = transforms.ToTensor() 20 | 21 | train_dataset = datasets.CIFAR10(root='data', 22 | train=True, 23 | transform=train_transforms, 24 | download=download) 25 | 26 | valid_dataset = datasets.CIFAR10(root='data', 27 | train=True, 28 | transform=test_transforms) 29 | 30 | test_dataset = datasets.CIFAR10(root='data', 31 | train=False, 32 | transform=test_transforms) 33 | 34 | if validation_fraction is not None: 35 | num = int(validation_fraction * 50000) 36 | train_indices = range(0, 50000 - num) 37 | valid_indices = range(50000 - num, 50000) 38 | 39 | train_sampler = SubsetRandomSampler(train_indices) 40 | valid_sampler = SubsetRandomSampler(valid_indices) 41 | 42 | valid_loader = DataLoader(dataset=valid_dataset, 43 | batch_size=batch_size, 44 | num_workers=num_workers, 45 | sampler=valid_sampler) 46 | 47 | train_loader = DataLoader(dataset=train_dataset, 48 | batch_size=batch_size, 49 | num_workers=num_workers, 50 | drop_last=True, 51 | sampler=train_sampler) 52 | 53 | else: 54 | train_loader = DataLoader(dataset=train_dataset, 55 | batch_size=batch_size, 56 | num_workers=num_workers, 57 | drop_last=True, 58 | shuffle=True) 59 | 60 | test_loader = DataLoader(dataset=test_dataset, 61 | batch_size=batch_size, 62 | num_workers=num_workers, 63 | shuffle=False) 64 | 65 | if validation_fraction is None: 66 | return train_loader, test_loader 67 | else: 68 | return train_loader, valid_loader, test_loader 69 | -------------------------------------------------------------------------------- /precision-benchmarks/torchvision-vit16-small/local_utilities.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from torch.utils.data import sampler 3 | from torchvision import datasets 4 | from torch.utils.data import DataLoader 5 | from torch.utils.data import SubsetRandomSampler 6 | from torchvision import transforms 7 | 8 | 9 | def get_dataloaders_cifar10(batch_size, num_workers=0, 10 | validation_fraction=None, 11 | train_transforms=None, 12 | test_transforms=None, 13 | download=True): 14 | 15 | if train_transforms is None: 16 | train_transforms = transforms.ToTensor() 17 | 18 | if test_transforms is None: 19 | test_transforms = transforms.ToTensor() 20 | 21 | train_dataset = datasets.CIFAR10(root='data', 22 | train=True, 23 | transform=train_transforms, 24 | download=download) 25 | 26 | valid_dataset = datasets.CIFAR10(root='data', 27 | train=True, 28 | transform=test_transforms) 29 | 30 | test_dataset = datasets.CIFAR10(root='data', 31 | train=False, 32 | transform=test_transforms) 33 | 34 | if validation_fraction is not None: 35 | num = int(validation_fraction * 50000) 36 | train_indices = range(0, 50000 - num) 37 | valid_indices = range(50000 - num, 50000) 38 | 39 | train_sampler = SubsetRandomSampler(train_indices) 40 | valid_sampler = SubsetRandomSampler(valid_indices) 41 | 42 | valid_loader = DataLoader(dataset=valid_dataset, 43 | batch_size=batch_size, 44 | num_workers=num_workers, 45 | sampler=valid_sampler) 46 | 47 | train_loader = DataLoader(dataset=train_dataset, 48 | batch_size=batch_size, 49 | num_workers=num_workers, 50 | drop_last=True, 51 | sampler=train_sampler) 52 | 53 | else: 54 | train_loader = DataLoader(dataset=train_dataset, 55 | batch_size=batch_size, 56 | num_workers=num_workers, 57 | drop_last=True, 58 | shuffle=True) 59 | 60 | test_loader = DataLoader(dataset=test_dataset, 61 | batch_size=batch_size, 62 | num_workers=num_workers, 63 | shuffle=False) 64 | 65 | if validation_fraction is None: 66 | return train_loader, test_loader 67 | else: 68 | return train_loader, valid_loader, test_loader 69 | -------------------------------------------------------------------------------- /.gitignore: -------------------------------------------------------------------------------- 1 | .DS_Store 2 | # Byte-compiled / optimized / DLL files 3 | __pycache__/ 4 | *.py[cod] 5 | *$py.class 6 | 7 | # C extensions 8 | *.so 9 | 10 | # Distribution / packaging 11 | .Python 12 | build/ 13 | develop-eggs/ 14 | dist/ 15 | downloads/ 16 | eggs/ 17 | .eggs/ 18 | lib/ 19 | lib64/ 20 | parts/ 21 | sdist/ 22 | var/ 23 | wheels/ 24 | share/python-wheels/ 25 | *.egg-info/ 26 | .installed.cfg 27 | *.egg 28 | MANIFEST 29 | 30 | # PyInstaller 31 | # Usually these files are written by a python script from a template 32 | # before PyInstaller builds the exe, so as to inject date/other infos into it. 33 | *.manifest 34 | *.spec 35 | 36 | # Installer logs 37 | pip-log.txt 38 | pip-delete-this-directory.txt 39 | 40 | # Unit test / coverage reports 41 | htmlcov/ 42 | .tox/ 43 | .nox/ 44 | .coverage 45 | .coverage.* 46 | .cache 47 | nosetests.xml 48 | coverage.xml 49 | *.cover 50 | *.py,cover 51 | .hypothesis/ 52 | .pytest_cache/ 53 | cover/ 54 | 55 | # Translations 56 | *.mo 57 | *.pot 58 | 59 | # Django stuff: 60 | *.log 61 | local_settings.py 62 | db.sqlite3 63 | db.sqlite3-journal 64 | 65 | # Flask stuff: 66 | instance/ 67 | .webassets-cache 68 | 69 | # Scrapy stuff: 70 | .scrapy 71 | 72 | # Sphinx documentation 73 | docs/_build/ 74 | 75 | # PyBuilder 76 | .pybuilder/ 77 | target/ 78 | 79 | # Jupyter Notebook 80 | .ipynb_checkpoints 81 | 82 | # IPython 83 | profile_default/ 84 | ipython_config.py 85 | 86 | # pyenv 87 | # For a library or package, you might want to ignore these files since the code is 88 | # intended to run in multiple environments; otherwise, check them in: 89 | # .python-version 90 | 91 | # pipenv 92 | # According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control. 93 | # However, in case of collaboration, if having platform-specific dependencies or dependencies 94 | # having no cross-platform support, pipenv may install dependencies that don't work, or not 95 | # install all needed dependencies. 96 | #Pipfile.lock 97 | 98 | # poetry 99 | # Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control. 100 | # This is especially recommended for binary packages to ensure reproducibility, and is more 101 | # commonly ignored for libraries. 102 | # https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control 103 | #poetry.lock 104 | 105 | # pdm 106 | # Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control. 107 | #pdm.lock 108 | # pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it 109 | # in version control. 110 | # https://pdm.fming.dev/#use-with-ide 111 | .pdm.toml 112 | 113 | # PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm 114 | __pypackages__/ 115 | 116 | # Celery stuff 117 | celerybeat-schedule 118 | celerybeat.pid 119 | 120 | # SageMath parsed files 121 | *.sage.py 122 | 123 | # Environments 124 | .env 125 | .venv 126 | env/ 127 | venv/ 128 | ENV/ 129 | env.bak/ 130 | venv.bak/ 131 | 132 | # Spyder project settings 133 | .spyderproject 134 | .spyproject 135 | 136 | # Rope project settings 137 | .ropeproject 138 | 139 | # mkdocs documentation 140 | /site 141 | 142 | # mypy 143 | .mypy_cache/ 144 | .dmypy.json 145 | dmypy.json 146 | 147 | # Pyre type checker 148 | .pyre/ 149 | 150 | # pytype static type analyzer 151 | .pytype/ 152 | 153 | # Cython debug symbols 154 | cython_debug/ 155 | 156 | # PyCharm 157 | # JetBrains specific template is maintained in a separate JetBrains.gitignore that can 158 | # be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore 159 | # and can be added to the global gitignore or merged into this file. For a more nuclear 160 | # option (not recommended) you can uncomment the following to ignore the entire idea folder. 161 | #.idea/ 162 | -------------------------------------------------------------------------------- /precision-benchmarks/torchvision-vit16-large/trainer_utilities.py: -------------------------------------------------------------------------------- 1 | import lightning as L 2 | import torch 3 | import torch.nn.functional as F 4 | import torchmetrics 5 | from torchvision import datasets 6 | from torch.utils.data import DataLoader 7 | from torch.utils.data import SubsetRandomSampler 8 | from torchvision import transforms 9 | 10 | 11 | class LightningModel(L.LightningModule): 12 | def __init__(self, model, learning_rate): 13 | super().__init__() 14 | 15 | self.learning_rate = learning_rate 16 | self.model = model 17 | 18 | self.save_hyperparameters(ignore=["model"]) 19 | 20 | self.train_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10) 21 | self.val_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10) 22 | self.test_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10) 23 | 24 | def forward(self, x): 25 | return self.model(x) 26 | 27 | def _shared_step(self, batch): 28 | features, true_labels = batch 29 | logits = self(features) 30 | 31 | loss = F.cross_entropy(logits, true_labels) 32 | predicted_labels = torch.argmax(logits, dim=1) 33 | return loss, true_labels, predicted_labels 34 | 35 | def training_step(self, batch, batch_idx): 36 | loss, true_labels, predicted_labels = self._shared_step(batch) 37 | 38 | self.log("train_loss", loss) 39 | self.train_acc(predicted_labels, true_labels) 40 | self.log( 41 | "train_acc", self.train_acc, prog_bar=True, on_epoch=True, on_step=False 42 | ) 43 | return loss 44 | 45 | def validation_step(self, batch, batch_idx): 46 | loss, true_labels, predicted_labels = self._shared_step(batch) 47 | 48 | self.log("val_loss", loss, prog_bar=True) 49 | self.val_acc(predicted_labels, true_labels) 50 | self.log("val_acc", self.val_acc, prog_bar=True) 51 | 52 | def test_step(self, batch, batch_idx): 53 | loss, true_labels, predicted_labels = self._shared_step(batch) 54 | self.test_acc(predicted_labels, true_labels) 55 | self.log("test_acc", self.test_acc) 56 | 57 | def configure_optimizers(self): 58 | optimizer = torch.optim.Adam(self.parameters(), lr=self.learning_rate) 59 | return optimizer 60 | 61 | 62 | def get_dataloaders_cifar10(batch_size, num_workers=0, 63 | validation_fraction=None, 64 | train_transforms=None, 65 | test_transforms=None, 66 | download=True): 67 | 68 | if train_transforms is None: 69 | train_transforms = transforms.ToTensor() 70 | 71 | if test_transforms is None: 72 | test_transforms = transforms.ToTensor() 73 | 74 | train_dataset = datasets.CIFAR10(root='data', 75 | train=True, 76 | transform=train_transforms, 77 | download=download) 78 | 79 | valid_dataset = datasets.CIFAR10(root='data', 80 | train=True, 81 | transform=test_transforms) 82 | 83 | test_dataset = datasets.CIFAR10(root='data', 84 | train=False, 85 | transform=test_transforms) 86 | 87 | if validation_fraction is not None: 88 | num = int(validation_fraction * 50000) 89 | train_indices = range(0, 50000 - num) 90 | valid_indices = range(50000 - num, 50000) 91 | 92 | train_sampler = SubsetRandomSampler(train_indices) 93 | valid_sampler = SubsetRandomSampler(valid_indices) 94 | 95 | valid_loader = DataLoader(dataset=valid_dataset, 96 | batch_size=batch_size, 97 | num_workers=num_workers, 98 | sampler=valid_sampler) 99 | 100 | train_loader = DataLoader(dataset=train_dataset, 101 | batch_size=batch_size, 102 | num_workers=num_workers, 103 | drop_last=True, 104 | sampler=train_sampler) 105 | 106 | else: 107 | train_loader = DataLoader(dataset=train_dataset, 108 | batch_size=batch_size, 109 | num_workers=num_workers, 110 | drop_last=True, 111 | shuffle=True) 112 | 113 | test_loader = DataLoader(dataset=test_dataset, 114 | batch_size=batch_size, 115 | num_workers=num_workers, 116 | shuffle=False) 117 | 118 | if validation_fraction is None: 119 | return train_loader, test_loader 120 | else: 121 | return train_loader, valid_loader, test_loader 122 | -------------------------------------------------------------------------------- /precision-benchmarks/torchvision-vit16-small/trainer_utilities.py: -------------------------------------------------------------------------------- 1 | import lightning as L 2 | import torch 3 | import torch.nn.functional as F 4 | import torchmetrics 5 | from torchvision import datasets 6 | from torch.utils.data import DataLoader 7 | from torch.utils.data import SubsetRandomSampler 8 | from torchvision import transforms 9 | 10 | 11 | class LightningModel(L.LightningModule): 12 | def __init__(self, model, learning_rate): 13 | super().__init__() 14 | 15 | self.learning_rate = learning_rate 16 | self.model = model 17 | 18 | self.save_hyperparameters(ignore=["model"]) 19 | 20 | self.train_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10) 21 | self.val_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10) 22 | self.test_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10) 23 | 24 | def forward(self, x): 25 | return self.model(x) 26 | 27 | def _shared_step(self, batch): 28 | features, true_labels = batch 29 | logits = self(features) 30 | 31 | loss = F.cross_entropy(logits, true_labels) 32 | predicted_labels = torch.argmax(logits, dim=1) 33 | return loss, true_labels, predicted_labels 34 | 35 | def training_step(self, batch, batch_idx): 36 | loss, true_labels, predicted_labels = self._shared_step(batch) 37 | 38 | self.log("train_loss", loss) 39 | self.train_acc(predicted_labels, true_labels) 40 | self.log( 41 | "train_acc", self.train_acc, prog_bar=True, on_epoch=True, on_step=False 42 | ) 43 | return loss 44 | 45 | def validation_step(self, batch, batch_idx): 46 | loss, true_labels, predicted_labels = self._shared_step(batch) 47 | 48 | self.log("val_loss", loss, prog_bar=True) 49 | self.val_acc(predicted_labels, true_labels) 50 | self.log("val_acc", self.val_acc, prog_bar=True) 51 | 52 | def test_step(self, batch, batch_idx): 53 | loss, true_labels, predicted_labels = self._shared_step(batch) 54 | self.test_acc(predicted_labels, true_labels) 55 | self.log("test_acc", self.test_acc) 56 | 57 | def configure_optimizers(self): 58 | optimizer = torch.optim.Adam(self.parameters(), lr=self.learning_rate) 59 | return optimizer 60 | 61 | 62 | def get_dataloaders_cifar10(batch_size, num_workers=0, 63 | validation_fraction=None, 64 | train_transforms=None, 65 | test_transforms=None, 66 | download=True): 67 | 68 | if train_transforms is None: 69 | train_transforms = transforms.ToTensor() 70 | 71 | if test_transforms is None: 72 | test_transforms = transforms.ToTensor() 73 | 74 | train_dataset = datasets.CIFAR10(root='data', 75 | train=True, 76 | transform=train_transforms, 77 | download=download) 78 | 79 | valid_dataset = datasets.CIFAR10(root='data', 80 | train=True, 81 | transform=test_transforms) 82 | 83 | test_dataset = datasets.CIFAR10(root='data', 84 | train=False, 85 | transform=test_transforms) 86 | 87 | if validation_fraction is not None: 88 | num = int(validation_fraction * 50000) 89 | train_indices = range(0, 50000 - num) 90 | valid_indices = range(50000 - num, 50000) 91 | 92 | train_sampler = SubsetRandomSampler(train_indices) 93 | valid_sampler = SubsetRandomSampler(valid_indices) 94 | 95 | valid_loader = DataLoader(dataset=valid_dataset, 96 | batch_size=batch_size, 97 | num_workers=num_workers, 98 | sampler=valid_sampler) 99 | 100 | train_loader = DataLoader(dataset=train_dataset, 101 | batch_size=batch_size, 102 | num_workers=num_workers, 103 | drop_last=True, 104 | sampler=train_sampler) 105 | 106 | else: 107 | train_loader = DataLoader(dataset=train_dataset, 108 | batch_size=batch_size, 109 | num_workers=num_workers, 110 | drop_last=True, 111 | shuffle=True) 112 | 113 | test_loader = DataLoader(dataset=test_dataset, 114 | batch_size=batch_size, 115 | num_workers=num_workers, 116 | shuffle=False) 117 | 118 | if validation_fraction is None: 119 | return train_loader, test_loader 120 | else: 121 | return train_loader, valid_loader, test_loader 122 | -------------------------------------------------------------------------------- /precision-benchmarks/torchvision-vit16-large/01_pytorch-fp32.py: -------------------------------------------------------------------------------- 1 | import time 2 | 3 | import torch 4 | import torch.nn.functional as F 5 | from torch.optim.lr_scheduler import ExponentialLR 6 | import torchmetrics 7 | from torchvision import transforms 8 | from torchvision.models import vit_l_16 9 | from torchvision.models import ViT_L_16_Weights 10 | 11 | from local_utilities import get_dataloaders_cifar10 12 | 13 | 14 | def train(num_epochs, model, optimizer, train_loader, val_loader, device, scheduler): 15 | 16 | for epoch in range(num_epochs): 17 | train_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(device) 18 | 19 | model.train() 20 | for batch_idx, (features, targets) in enumerate(train_loader): 21 | model.train() 22 | 23 | ### FORWARD AND BACK PROP 24 | features, targets = features.to(device), targets.to(device) 25 | logits = model(features) 26 | loss = F.cross_entropy(logits, targets) 27 | loss.backward() 28 | 29 | ### UPDATE MODEL PARAMETERS 30 | optimizer.step() 31 | optimizer.zero_grad() 32 | 33 | ### LOGGING 34 | if not batch_idx % 300: 35 | print(f"Epoch: {epoch+1:04d}/{num_epochs:04d} | Batch {batch_idx:04d}/{len(train_loader):04d} | Loss: {loss:.4f}") 36 | 37 | model.eval() 38 | with torch.no_grad(): 39 | predicted_labels = torch.argmax(logits, 1) 40 | train_acc.update(predicted_labels, targets) 41 | scheduler.step() 42 | 43 | ### MORE LOGGING 44 | model.eval() 45 | with torch.no_grad(): 46 | val_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(device) 47 | 48 | for (features, targets) in val_loader: 49 | features, targets = features.to(device), targets.to(device) 50 | outputs = model(features) 51 | predicted_labels = torch.argmax(outputs, 1) 52 | val_acc.update(predicted_labels, targets) 53 | 54 | print(f"Epoch: {epoch+1:04d}/{num_epochs:04d} | Train acc.: {train_acc.compute()*100:.2f}% | Val acc.: {val_acc.compute()*100:.2f}%") 55 | train_acc.reset(), val_acc.reset() 56 | 57 | 58 | if __name__ == "__main__": 59 | 60 | print("PyTorch:", torch.__version__) 61 | torch.set_float32_matmul_precision("medium") 62 | 63 | device = torch.device("cuda") 64 | 65 | ########################## 66 | ### 1 Loading the Dataset 67 | ########################## 68 | train_transforms = transforms.Compose([transforms.Resize((224, 224)), 69 | #transforms.RandomCrop((224, 224)), 70 | transforms.ToTensor()]) 71 | 72 | test_transforms = transforms.Compose([transforms.Resize((224, 224)), 73 | #transforms.CenterCrop((224, 224)), 74 | transforms.ToTensor()]) 75 | 76 | train_loader, val_loader, test_loader = get_dataloaders_cifar10( 77 | batch_size=32, 78 | num_workers=4, 79 | train_transforms=train_transforms, 80 | test_transforms=test_transforms, 81 | validation_fraction=0.1, 82 | download=True 83 | ) 84 | 85 | ######################################### 86 | ### 2 Initializing the Model 87 | ######################################### 88 | 89 | model = vit_l_16(weights=ViT_L_16_Weights.IMAGENET1K_V1) 90 | 91 | # replace output layer 92 | model.heads.head = torch.nn.Linear(in_features=1024, out_features=10) 93 | model.to(device) 94 | 95 | optimizer = torch.optim.Adam(model.parameters(), lr=5e-5) 96 | scheduler = ExponentialLR(optimizer, gamma=0.9) 97 | 98 | ######################################### 99 | ### 3 Finetuning 100 | ######################################### 101 | 102 | start = time.time() 103 | train( 104 | num_epochs=3, 105 | model=model, 106 | optimizer=optimizer, 107 | train_loader=train_loader, 108 | val_loader=val_loader, 109 | device=device, 110 | scheduler=scheduler 111 | ) 112 | 113 | end = time.time() 114 | elapsed = end-start 115 | print(f"Time elapsed {elapsed/60:.2f} min") 116 | print(f"Memory used: {torch.cuda.max_memory_reserved() / 1e9:.02f} GB") 117 | 118 | ######################################### 119 | ### 4 Evaluation 120 | ######################################### 121 | 122 | with torch.no_grad(): 123 | model.eval() 124 | test_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(device) 125 | 126 | for (features, targets) in test_loader: 127 | features, targets = features.to(device), targets.to(device) 128 | outputs = model(features) 129 | predicted_labels = torch.argmax(outputs, 1) 130 | test_acc.update(predicted_labels, targets) 131 | 132 | print(f"Test accuracy {test_acc.compute()*100:.2f}%") -------------------------------------------------------------------------------- /precision-benchmarks/torchvision-vit16-small/01_pytorch-fp32.py: -------------------------------------------------------------------------------- 1 | import time 2 | 3 | import torch 4 | import torch.nn.functional as F 5 | from torch.optim.lr_scheduler import ExponentialLR 6 | import torchmetrics 7 | from torchvision import transforms 8 | from torchvision.models import vit_b_16 9 | from torchvision.models import ViT_B_16_Weights 10 | 11 | from local_utilities import get_dataloaders_cifar10 12 | 13 | 14 | def train(num_epochs, model, optimizer, train_loader, val_loader, device, scheduler): 15 | 16 | for epoch in range(num_epochs): 17 | train_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(device) 18 | 19 | model.train() 20 | for batch_idx, (features, targets) in enumerate(train_loader): 21 | model.train() 22 | 23 | ### FORWARD AND BACK PROP 24 | features, targets = features.to(device), targets.to(device) 25 | logits = model(features) 26 | loss = F.cross_entropy(logits, targets) 27 | loss.backward() 28 | 29 | ### UPDATE MODEL PARAMETERS 30 | optimizer.step() 31 | optimizer.zero_grad() 32 | 33 | ### LOGGING 34 | if not batch_idx % 300: 35 | print(f"Epoch: {epoch+1:04d}/{num_epochs:04d} | Batch {batch_idx:04d}/{len(train_loader):04d} | Loss: {loss:.4f}") 36 | 37 | model.eval() 38 | with torch.no_grad(): 39 | predicted_labels = torch.argmax(logits, 1) 40 | train_acc.update(predicted_labels, targets) 41 | scheduler.step() 42 | 43 | ### MORE LOGGING 44 | model.eval() 45 | with torch.no_grad(): 46 | val_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(device) 47 | 48 | for (features, targets) in val_loader: 49 | features, targets = features.to(device), targets.to(device) 50 | outputs = model(features) 51 | predicted_labels = torch.argmax(outputs, 1) 52 | val_acc.update(predicted_labels, targets) 53 | 54 | print(f"Epoch: {epoch+1:04d}/{num_epochs:04d} | Train acc.: {train_acc.compute()*100:.2f}% | Val acc.: {val_acc.compute()*100:.2f}%") 55 | train_acc.reset(), val_acc.reset() 56 | 57 | 58 | if __name__ == "__main__": 59 | 60 | print("PyTorch:", torch.__version__) 61 | torch.set_float32_matmul_precision("medium") 62 | 63 | device = torch.device("cuda") 64 | 65 | ########################## 66 | ### 1 Loading the Dataset 67 | ########################## 68 | train_transforms = transforms.Compose([transforms.Resize((224, 224)), 69 | #transforms.RandomCrop((224, 224)), 70 | transforms.ToTensor()]) 71 | 72 | test_transforms = transforms.Compose([transforms.Resize((224, 224)), 73 | #transforms.CenterCrop((224, 224)), 74 | transforms.ToTensor()]) 75 | 76 | train_loader, val_loader, test_loader = get_dataloaders_cifar10( 77 | batch_size=16, 78 | num_workers=4, 79 | train_transforms=train_transforms, 80 | test_transforms=test_transforms, 81 | validation_fraction=0.1, 82 | download=True 83 | ) 84 | 85 | ######################################### 86 | ### 2 Initializing the Model 87 | ######################################### 88 | 89 | model = vit_b_16(weights=ViT_B_16_Weights.IMAGENET1K_V1) 90 | 91 | # replace output layer 92 | model.heads.head = torch.nn.Linear(in_features=768, out_features=10) 93 | model.to(device) 94 | 95 | optimizer = torch.optim.Adam(model.parameters(), lr=5e-5) 96 | scheduler = ExponentialLR(optimizer, gamma=0.9) 97 | 98 | ######################################### 99 | ### 3 Finetuning 100 | ######################################### 101 | 102 | start = time.time() 103 | train( 104 | num_epochs=3, 105 | model=model, 106 | optimizer=optimizer, 107 | train_loader=train_loader, 108 | val_loader=val_loader, 109 | device=device, 110 | scheduler=scheduler 111 | ) 112 | 113 | end = time.time() 114 | elapsed = end-start 115 | print(f"Time elapsed {elapsed/60:.2f} min") 116 | print(f"Memory used: {torch.cuda.max_memory_reserved() / 1e9:.02f} GB") 117 | 118 | ######################################### 119 | ### 4 Evaluation 120 | ######################################### 121 | 122 | with torch.no_grad(): 123 | model.eval() 124 | test_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(device) 125 | 126 | for (features, targets) in test_loader: 127 | features, targets = features.to(device), targets.to(device) 128 | outputs = model(features) 129 | predicted_labels = torch.argmax(outputs, 1) 130 | test_acc.update(predicted_labels, targets) 131 | 132 | print(f"Test accuracy {test_acc.compute()*100:.2f}%") -------------------------------------------------------------------------------- /precision-benchmarks/torchvision-vit16-small/02_pytorch-fabric-fp32.py: -------------------------------------------------------------------------------- 1 | import time 2 | 3 | import lightning as L 4 | from lightning import Fabric 5 | import torch 6 | import torch.nn.functional as F 7 | from torch.optim.lr_scheduler import ExponentialLR 8 | import torchmetrics 9 | from torchvision import transforms 10 | from torchvision.models import vit_b_16 11 | from torchvision.models import ViT_B_16_Weights 12 | 13 | from local_utilities import get_dataloaders_cifar10 14 | 15 | 16 | def train(num_epochs, model, optimizer, train_loader, val_loader, fabric, scheduler): 17 | 18 | for epoch in range(num_epochs): 19 | train_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(fabric.device) 20 | 21 | model.train() 22 | for batch_idx, (features, targets) in enumerate(train_loader): 23 | model.train() 24 | 25 | ### FORWARD AND BACK PROP 26 | logits = model(features) 27 | loss = F.cross_entropy(logits, targets) 28 | fabric.backward(loss) 29 | 30 | ### UPDATE MODEL PARAMETERS 31 | optimizer.step() 32 | optimizer.zero_grad() 33 | 34 | ### LOGGING 35 | if not batch_idx % 300: 36 | fabric.print(f"Epoch: {epoch+1:04d}/{num_epochs:04d} | Batch {batch_idx:04d}/{len(train_loader):04d} | Loss: {loss:.4f}") 37 | 38 | model.eval() 39 | with torch.no_grad(): 40 | predicted_labels = torch.argmax(logits, 1) 41 | train_acc.update(predicted_labels, targets) 42 | scheduler.step() 43 | 44 | ### MORE LOGGING 45 | model.eval() 46 | with torch.no_grad(): 47 | val_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(fabric.device) 48 | 49 | for (features, targets) in val_loader: 50 | outputs = model(features) 51 | predicted_labels = torch.argmax(outputs, 1) 52 | val_acc.update(predicted_labels, targets) 53 | 54 | fabric.print(f"Epoch: {epoch+1:04d}/{num_epochs:04d} | Train acc.: {train_acc.compute()*100:.2f}% | Val acc.: {val_acc.compute()*100:.2f}%") 55 | train_acc.reset(), val_acc.reset() 56 | 57 | 58 | if __name__ == "__main__": 59 | 60 | print("PyTorch:", torch.__version__) 61 | print("Lightning:", L.__version__) 62 | torch.set_float32_matmul_precision("medium") 63 | 64 | L.seed_everything(123) 65 | 66 | ########################## 67 | ### 1 Loading the Dataset 68 | ########################## 69 | train_transforms = transforms.Compose([transforms.Resize((224, 224)), 70 | #transforms.RandomCrop((224, 224)), 71 | transforms.ToTensor()]) 72 | 73 | test_transforms = transforms.Compose([transforms.Resize((224, 224)), 74 | #transforms.CenterCrop((224, 224)), 75 | transforms.ToTensor()]) 76 | 77 | train_loader, val_loader, test_loader = get_dataloaders_cifar10( 78 | batch_size=16, 79 | num_workers=4, 80 | train_transforms=train_transforms, 81 | test_transforms=test_transforms, 82 | validation_fraction=0.1, 83 | download=True 84 | ) 85 | 86 | ######################################### 87 | ### 2 Initializing the Model 88 | ######################################### 89 | 90 | model = vit_b_16(weights=ViT_B_16_Weights.IMAGENET1K_V1) 91 | 92 | # replace output layer 93 | model.heads.head = torch.nn.Linear(in_features=768, out_features=10) 94 | 95 | optimizer = torch.optim.Adam(model.parameters(), lr=5e-5) 96 | scheduler = ExponentialLR(optimizer, gamma=0.9) 97 | 98 | ######################################### 99 | ### 3 Launch Fabric 100 | ######################################### 101 | 102 | fabric = Fabric(accelerator="cuda", devices=1) 103 | fabric.launch() 104 | 105 | train_loader, val_loader, test_loader = fabric.setup_dataloaders( 106 | train_loader, val_loader, test_loader) 107 | model, optimizer = fabric.setup(model, optimizer) 108 | 109 | ######################################### 110 | ### 4 Finetuning 111 | ######################################### 112 | 113 | start = time.time() 114 | train( 115 | num_epochs=3, 116 | model=model, 117 | optimizer=optimizer, 118 | train_loader=train_loader, 119 | val_loader=val_loader, 120 | fabric=fabric, 121 | scheduler=scheduler 122 | ) 123 | 124 | end = time.time() 125 | elapsed = end-start 126 | fabric.print(f"Time elapsed {elapsed/60:.2f} min") 127 | fabric.print(f"Memory used: {torch.cuda.max_memory_reserved() / 1e9:.02f} GB") 128 | 129 | ######################################### 130 | ### 5 Evaluation 131 | ######################################### 132 | 133 | with torch.no_grad(): 134 | model.eval() 135 | test_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(fabric.device) 136 | 137 | for (features, targets) in test_loader: 138 | outputs = model(features) 139 | predicted_labels = torch.argmax(outputs, 1) 140 | test_acc.update(predicted_labels, targets) 141 | 142 | fabric.print(f"Test accuracy {test_acc.compute()*100:.2f}%") -------------------------------------------------------------------------------- /precision-benchmarks/torchvision-vit16-large/02_pytorch-fabric-fp32.py: -------------------------------------------------------------------------------- 1 | import time 2 | 3 | import lightning as L 4 | from lightning import Fabric 5 | import torch 6 | import torch.nn.functional as F 7 | from torch.optim.lr_scheduler import ExponentialLR 8 | import torchmetrics 9 | from torchvision import transforms 10 | from torchvision.models import vit_l_16 11 | from torchvision.models import ViT_L_16_Weights 12 | 13 | from local_utilities import get_dataloaders_cifar10 14 | 15 | 16 | def train(num_epochs, model, optimizer, train_loader, val_loader, fabric, scheduler): 17 | 18 | for epoch in range(num_epochs): 19 | train_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(fabric.device) 20 | 21 | model.train() 22 | for batch_idx, (features, targets) in enumerate(train_loader): 23 | model.train() 24 | 25 | ### FORWARD AND BACK PROP 26 | logits = model(features) 27 | loss = F.cross_entropy(logits, targets) 28 | fabric.backward(loss) 29 | 30 | ### UPDATE MODEL PARAMETERS 31 | optimizer.step() 32 | optimizer.zero_grad() 33 | 34 | ### LOGGING 35 | if not batch_idx % 300: 36 | fabric.print(f"Epoch: {epoch+1:04d}/{num_epochs:04d} | Batch {batch_idx:04d}/{len(train_loader):04d} | Loss: {loss:.4f}") 37 | 38 | model.eval() 39 | with torch.no_grad(): 40 | predicted_labels = torch.argmax(logits, 1) 41 | train_acc.update(predicted_labels, targets) 42 | scheduler.step() 43 | 44 | ### MORE LOGGING 45 | model.eval() 46 | with torch.no_grad(): 47 | val_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(fabric.device) 48 | 49 | for (features, targets) in val_loader: 50 | outputs = model(features) 51 | predicted_labels = torch.argmax(outputs, 1) 52 | val_acc.update(predicted_labels, targets) 53 | 54 | fabric.print(f"Epoch: {epoch+1:04d}/{num_epochs:04d} | Train acc.: {train_acc.compute()*100:.2f}% | Val acc.: {val_acc.compute()*100:.2f}%") 55 | train_acc.reset(), val_acc.reset() 56 | 57 | 58 | if __name__ == "__main__": 59 | 60 | print("PyTorch:", torch.__version__) 61 | print("Lightning:", L.__version__) 62 | torch.set_float32_matmul_precision("medium") 63 | 64 | L.seed_everything(123) 65 | 66 | ########################## 67 | ### 1 Loading the Dataset 68 | ########################## 69 | train_transforms = transforms.Compose([transforms.Resize((224, 224)), 70 | #transforms.RandomCrop((224, 224)), 71 | transforms.ToTensor()]) 72 | 73 | test_transforms = transforms.Compose([transforms.Resize((224, 224)), 74 | #transforms.CenterCrop((224, 224)), 75 | transforms.ToTensor()]) 76 | 77 | train_loader, val_loader, test_loader = get_dataloaders_cifar10( 78 | batch_size=32, 79 | num_workers=4, 80 | train_transforms=train_transforms, 81 | test_transforms=test_transforms, 82 | validation_fraction=0.1, 83 | download=True 84 | ) 85 | 86 | ######################################### 87 | ### 2 Initializing the Model 88 | ######################################### 89 | 90 | model = vit_l_16(weights=ViT_L_16_Weights.IMAGENET1K_V1) 91 | 92 | # replace output layer 93 | model.heads.head = torch.nn.Linear(in_features=1024, out_features=10) 94 | 95 | optimizer = torch.optim.Adam(model.parameters(), lr=5e-5) 96 | scheduler = ExponentialLR(optimizer, gamma=0.9) 97 | 98 | ######################################### 99 | ### 3 Launch Fabric 100 | ######################################### 101 | 102 | fabric = Fabric(accelerator="cuda", devices=1) 103 | fabric.launch() 104 | 105 | train_loader, val_loader, test_loader = fabric.setup_dataloaders( 106 | train_loader, val_loader, test_loader) 107 | model, optimizer = fabric.setup(model, optimizer) 108 | 109 | ######################################### 110 | ### 4 Finetuning 111 | ######################################### 112 | 113 | start = time.time() 114 | train( 115 | num_epochs=3, 116 | model=model, 117 | optimizer=optimizer, 118 | train_loader=train_loader, 119 | val_loader=val_loader, 120 | fabric=fabric, 121 | scheduler=scheduler 122 | ) 123 | 124 | end = time.time() 125 | elapsed = end-start 126 | fabric.print(f"Time elapsed {elapsed/60:.2f} min") 127 | fabric.print(f"Memory used: {torch.cuda.max_memory_reserved() / 1e9:.02f} GB") 128 | 129 | ######################################### 130 | ### 5 Evaluation 131 | ######################################### 132 | 133 | with torch.no_grad(): 134 | model.eval() 135 | test_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(fabric.device) 136 | 137 | for (features, targets) in test_loader: 138 | outputs = model(features) 139 | predicted_labels = torch.argmax(outputs, 1) 140 | test_acc.update(predicted_labels, targets) 141 | 142 | fabric.print(f"Test accuracy {test_acc.compute()*100:.2f}%") -------------------------------------------------------------------------------- /precision-benchmarks/torchvision-vit16-large/03_fp16-mixed.py: -------------------------------------------------------------------------------- 1 | import time 2 | 3 | import lightning as L 4 | from lightning import Fabric 5 | import torch 6 | import torch.nn.functional as F 7 | from torch.optim.lr_scheduler import ExponentialLR 8 | import torchmetrics 9 | from torchvision import transforms 10 | from torchvision.models import vit_l_16 11 | from torchvision.models import ViT_L_16_Weights 12 | 13 | from local_utilities import get_dataloaders_cifar10 14 | 15 | 16 | def train(num_epochs, model, optimizer, train_loader, val_loader, fabric, scheduler): 17 | 18 | for epoch in range(num_epochs): 19 | train_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(fabric.device) 20 | 21 | model.train() 22 | for batch_idx, (features, targets) in enumerate(train_loader): 23 | model.train() 24 | 25 | ### FORWARD AND BACK PROP 26 | logits = model(features) 27 | loss = F.cross_entropy(logits, targets) 28 | fabric.backward(loss) 29 | 30 | ### UPDATE MODEL PARAMETERS 31 | optimizer.step() 32 | optimizer.zero_grad() 33 | 34 | ### LOGGING 35 | if not batch_idx % 300: 36 | fabric.print(f"Epoch: {epoch+1:04d}/{num_epochs:04d} | Batch {batch_idx:04d}/{len(train_loader):04d} | Loss: {loss:.4f}") 37 | 38 | model.eval() 39 | with torch.no_grad(): 40 | predicted_labels = torch.argmax(logits, 1) 41 | train_acc.update(predicted_labels, targets) 42 | scheduler.step() 43 | 44 | ### MORE LOGGING 45 | model.eval() 46 | with torch.no_grad(): 47 | val_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(fabric.device) 48 | 49 | for (features, targets) in val_loader: 50 | outputs = model(features) 51 | predicted_labels = torch.argmax(outputs, 1) 52 | val_acc.update(predicted_labels, targets) 53 | 54 | fabric.print(f"Epoch: {epoch+1:04d}/{num_epochs:04d} | Train acc.: {train_acc.compute()*100:.2f}% | Val acc.: {val_acc.compute()*100:.2f}%") 55 | train_acc.reset(), val_acc.reset() 56 | 57 | 58 | if __name__ == "__main__": 59 | 60 | print("PyTorch:", torch.__version__) 61 | print("Lightning:", L.__version__) 62 | torch.set_float32_matmul_precision("medium") 63 | 64 | L.seed_everything(123) 65 | 66 | ########################## 67 | ### 1 Loading the Dataset 68 | ########################## 69 | train_transforms = transforms.Compose([transforms.Resize((224, 224)), 70 | #transforms.RandomCrop((224, 224)), 71 | transforms.ToTensor()]) 72 | 73 | test_transforms = transforms.Compose([transforms.Resize((224, 224)), 74 | #transforms.CenterCrop((224, 224)), 75 | transforms.ToTensor()]) 76 | 77 | train_loader, val_loader, test_loader = get_dataloaders_cifar10( 78 | batch_size=32, 79 | num_workers=4, 80 | train_transforms=train_transforms, 81 | test_transforms=test_transforms, 82 | validation_fraction=0.1, 83 | download=True 84 | ) 85 | 86 | ######################################### 87 | ### 2 Initializing the Model 88 | ######################################### 89 | 90 | model = vit_l_16(weights=ViT_L_16_Weights.IMAGENET1K_V1) 91 | 92 | # replace output layer 93 | model.heads.head = torch.nn.Linear(in_features=1024, out_features=10) 94 | 95 | optimizer = torch.optim.Adam(model.parameters(), lr=5e-5) 96 | scheduler = ExponentialLR(optimizer, gamma=0.9) 97 | 98 | ######################################### 99 | ### 3 Launch Fabric 100 | ######################################### 101 | 102 | fabric = Fabric(accelerator="cuda", devices=1, precision="16-mixed") 103 | fabric.launch() 104 | 105 | train_loader, val_loader, test_loader = fabric.setup_dataloaders( 106 | train_loader, val_loader, test_loader) 107 | model, optimizer = fabric.setup(model, optimizer) 108 | 109 | ######################################### 110 | ### 4 Finetuning 111 | ######################################### 112 | 113 | start = time.time() 114 | train( 115 | num_epochs=3, 116 | model=model, 117 | optimizer=optimizer, 118 | train_loader=train_loader, 119 | val_loader=val_loader, 120 | fabric=fabric, 121 | scheduler=scheduler 122 | ) 123 | 124 | end = time.time() 125 | elapsed = end-start 126 | fabric.print(f"Time elapsed {elapsed/60:.2f} min") 127 | fabric.print(f"Memory used: {torch.cuda.max_memory_reserved() / 1e9:.02f} GB") 128 | 129 | ######################################### 130 | ### 5 Evaluation 131 | ######################################### 132 | 133 | with torch.no_grad(): 134 | model.eval() 135 | test_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(fabric.device) 136 | 137 | for (features, targets) in test_loader: 138 | outputs = model(features) 139 | predicted_labels = torch.argmax(outputs, 1) 140 | test_acc.update(predicted_labels, targets) 141 | 142 | fabric.print(f"Test accuracy {test_acc.compute()*100:.2f}%") -------------------------------------------------------------------------------- /precision-benchmarks/torchvision-vit16-large/05_fp16-full.py: -------------------------------------------------------------------------------- 1 | import time 2 | 3 | import lightning as L 4 | from lightning import Fabric 5 | import torch 6 | import torch.nn.functional as F 7 | from torch.optim.lr_scheduler import ExponentialLR 8 | import torchmetrics 9 | from torchvision import transforms 10 | from torchvision.models import vit_l_16 11 | from torchvision.models import ViT_L_16_Weights 12 | 13 | from local_utilities import get_dataloaders_cifar10 14 | 15 | 16 | def train(num_epochs, model, optimizer, train_loader, val_loader, fabric, scheduler): 17 | 18 | for epoch in range(num_epochs): 19 | train_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(fabric.device) 20 | 21 | model.train() 22 | for batch_idx, (features, targets) in enumerate(train_loader): 23 | model.train() 24 | 25 | ### FORWARD AND BACK PROP 26 | logits = model(features) 27 | loss = F.cross_entropy(logits, targets) 28 | fabric.backward(loss) 29 | 30 | ### UPDATE MODEL PARAMETERS 31 | optimizer.step() 32 | optimizer.zero_grad() 33 | 34 | ### LOGGING 35 | if not batch_idx % 300: 36 | fabric.print(f"Epoch: {epoch+1:04d}/{num_epochs:04d} | Batch {batch_idx:04d}/{len(train_loader):04d} | Loss: {loss:.4f}") 37 | 38 | model.eval() 39 | with torch.no_grad(): 40 | predicted_labels = torch.argmax(logits, 1) 41 | train_acc.update(predicted_labels, targets) 42 | scheduler.step() 43 | 44 | ### MORE LOGGING 45 | model.eval() 46 | with torch.no_grad(): 47 | val_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(fabric.device) 48 | 49 | for (features, targets) in val_loader: 50 | outputs = model(features) 51 | predicted_labels = torch.argmax(outputs, 1) 52 | val_acc.update(predicted_labels, targets) 53 | 54 | fabric.print(f"Epoch: {epoch+1:04d}/{num_epochs:04d} | Train acc.: {train_acc.compute()*100:.2f}% | Val acc.: {val_acc.compute()*100:.2f}%") 55 | train_acc.reset(), val_acc.reset() 56 | 57 | 58 | if __name__ == "__main__": 59 | 60 | print("PyTorch:", torch.__version__) 61 | print("Lightning:", L.__version__) 62 | torch.set_float32_matmul_precision("medium") 63 | 64 | L.seed_everything(123) 65 | 66 | ########################## 67 | ### 1 Loading the Dataset 68 | ########################## 69 | train_transforms = transforms.Compose([transforms.Resize((224, 224)), 70 | #transforms.RandomCrop((224, 224)), 71 | transforms.ToTensor()]) 72 | 73 | test_transforms = transforms.Compose([transforms.Resize((224, 224)), 74 | #transforms.CenterCrop((224, 224)), 75 | transforms.ToTensor()]) 76 | 77 | train_loader, val_loader, test_loader = get_dataloaders_cifar10( 78 | batch_size=32, 79 | num_workers=4, 80 | train_transforms=train_transforms, 81 | test_transforms=test_transforms, 82 | validation_fraction=0.1, 83 | download=True 84 | ) 85 | 86 | ######################################### 87 | ### 2 Initializing the Model 88 | ######################################### 89 | 90 | model = vit_l_16(weights=ViT_L_16_Weights.IMAGENET1K_V1) 91 | 92 | # replace output layer 93 | model.heads.head = torch.nn.Linear(in_features=1024, out_features=10) 94 | 95 | optimizer = torch.optim.Adam(model.parameters(), lr=5e-5) 96 | scheduler = ExponentialLR(optimizer, gamma=0.9) 97 | 98 | ######################################### 99 | ### 3 Launch Fabric 100 | ######################################### 101 | 102 | fabric = Fabric(accelerator="cuda", devices=1, precision="16-true") 103 | fabric.launch() 104 | 105 | train_loader, val_loader, test_loader = fabric.setup_dataloaders( 106 | train_loader, val_loader, test_loader) 107 | model, optimizer = fabric.setup(model, optimizer) 108 | 109 | ######################################### 110 | ### 4 Finetuning 111 | ######################################### 112 | 113 | start = time.time() 114 | train( 115 | num_epochs=3, 116 | model=model, 117 | optimizer=optimizer, 118 | train_loader=train_loader, 119 | val_loader=val_loader, 120 | fabric=fabric, 121 | scheduler=scheduler 122 | ) 123 | 124 | end = time.time() 125 | elapsed = end-start 126 | fabric.print(f"Time elapsed {elapsed/60:.2f} min") 127 | fabric.print(f"Memory used: {torch.cuda.max_memory_reserved() / 1e9:.02f} GB") 128 | 129 | ######################################### 130 | ### 5 Evaluation 131 | ######################################### 132 | 133 | with torch.no_grad(): 134 | model.eval() 135 | test_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(fabric.device) 136 | 137 | for (features, targets) in test_loader: 138 | outputs = model(features) 139 | predicted_labels = torch.argmax(outputs, 1) 140 | test_acc.update(predicted_labels, targets) 141 | 142 | fabric.print(f"Test accuracy {test_acc.compute()*100:.2f}%") -------------------------------------------------------------------------------- /precision-benchmarks/torchvision-vit16-large/06_bf16-full.py: -------------------------------------------------------------------------------- 1 | import time 2 | 3 | import lightning as L 4 | from lightning import Fabric 5 | import torch 6 | import torch.nn.functional as F 7 | from torch.optim.lr_scheduler import ExponentialLR 8 | import torchmetrics 9 | from torchvision import transforms 10 | from torchvision.models import vit_l_16 11 | from torchvision.models import ViT_L_16_Weights 12 | 13 | from local_utilities import get_dataloaders_cifar10 14 | 15 | 16 | def train(num_epochs, model, optimizer, train_loader, val_loader, fabric, scheduler): 17 | 18 | for epoch in range(num_epochs): 19 | train_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(fabric.device) 20 | 21 | model.train() 22 | for batch_idx, (features, targets) in enumerate(train_loader): 23 | model.train() 24 | 25 | ### FORWARD AND BACK PROP 26 | logits = model(features) 27 | loss = F.cross_entropy(logits, targets) 28 | fabric.backward(loss) 29 | 30 | ### UPDATE MODEL PARAMETERS 31 | optimizer.step() 32 | optimizer.zero_grad() 33 | 34 | ### LOGGING 35 | if not batch_idx % 300: 36 | fabric.print(f"Epoch: {epoch+1:04d}/{num_epochs:04d} | Batch {batch_idx:04d}/{len(train_loader):04d} | Loss: {loss:.4f}") 37 | 38 | model.eval() 39 | with torch.no_grad(): 40 | predicted_labels = torch.argmax(logits, 1) 41 | train_acc.update(predicted_labels, targets) 42 | scheduler.step() 43 | 44 | ### MORE LOGGING 45 | model.eval() 46 | with torch.no_grad(): 47 | val_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(fabric.device) 48 | 49 | for (features, targets) in val_loader: 50 | outputs = model(features) 51 | predicted_labels = torch.argmax(outputs, 1) 52 | val_acc.update(predicted_labels, targets) 53 | 54 | fabric.print(f"Epoch: {epoch+1:04d}/{num_epochs:04d} | Train acc.: {train_acc.compute()*100:.2f}% | Val acc.: {val_acc.compute()*100:.2f}%") 55 | train_acc.reset(), val_acc.reset() 56 | 57 | 58 | if __name__ == "__main__": 59 | 60 | print("PyTorch:", torch.__version__) 61 | print("Lightning:", L.__version__) 62 | torch.set_float32_matmul_precision("medium") 63 | 64 | L.seed_everything(123) 65 | 66 | ########################## 67 | ### 1 Loading the Dataset 68 | ########################## 69 | train_transforms = transforms.Compose([transforms.Resize((224, 224)), 70 | #transforms.RandomCrop((224, 224)), 71 | transforms.ToTensor()]) 72 | 73 | test_transforms = transforms.Compose([transforms.Resize((224, 224)), 74 | #transforms.CenterCrop((224, 224)), 75 | transforms.ToTensor()]) 76 | 77 | train_loader, val_loader, test_loader = get_dataloaders_cifar10( 78 | batch_size=32, 79 | num_workers=4, 80 | train_transforms=train_transforms, 81 | test_transforms=test_transforms, 82 | validation_fraction=0.1, 83 | download=True 84 | ) 85 | 86 | ######################################### 87 | ### 2 Initializing the Model 88 | ######################################### 89 | 90 | model = vit_l_16(weights=ViT_L_16_Weights.IMAGENET1K_V1) 91 | 92 | # replace output layer 93 | model.heads.head = torch.nn.Linear(in_features=1024, out_features=10) 94 | 95 | optimizer = torch.optim.Adam(model.parameters(), lr=5e-5) 96 | scheduler = ExponentialLR(optimizer, gamma=0.9) 97 | 98 | ######################################### 99 | ### 3 Launch Fabric 100 | ######################################### 101 | 102 | fabric = Fabric(accelerator="cuda", devices=1, precision="bf16-true") 103 | fabric.launch() 104 | 105 | train_loader, val_loader, test_loader = fabric.setup_dataloaders( 106 | train_loader, val_loader, test_loader) 107 | model, optimizer = fabric.setup(model, optimizer) 108 | 109 | ######################################### 110 | ### 4 Finetuning 111 | ######################################### 112 | 113 | start = time.time() 114 | train( 115 | num_epochs=3, 116 | model=model, 117 | optimizer=optimizer, 118 | train_loader=train_loader, 119 | val_loader=val_loader, 120 | fabric=fabric, 121 | scheduler=scheduler 122 | ) 123 | 124 | end = time.time() 125 | elapsed = end-start 126 | fabric.print(f"Time elapsed {elapsed/60:.2f} min") 127 | fabric.print(f"Memory used: {torch.cuda.max_memory_reserved() / 1e9:.02f} GB") 128 | 129 | ######################################### 130 | ### 5 Evaluation 131 | ######################################### 132 | 133 | with torch.no_grad(): 134 | model.eval() 135 | test_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(fabric.device) 136 | 137 | for (features, targets) in test_loader: 138 | outputs = model(features) 139 | predicted_labels = torch.argmax(outputs, 1) 140 | test_acc.update(predicted_labels, targets) 141 | 142 | fabric.print(f"Test accuracy {test_acc.compute()*100:.2f}%") -------------------------------------------------------------------------------- /precision-benchmarks/torchvision-vit16-small/03_fp16-mixed.py: -------------------------------------------------------------------------------- 1 | import time 2 | 3 | import lightning as L 4 | from lightning import Fabric 5 | import torch 6 | import torch.nn.functional as F 7 | from torch.optim.lr_scheduler import ExponentialLR 8 | import torchmetrics 9 | from torchvision import transforms 10 | from torchvision.models import vit_b_16 11 | from torchvision.models import ViT_B_16_Weights 12 | 13 | from local_utilities import get_dataloaders_cifar10 14 | 15 | 16 | def train(num_epochs, model, optimizer, train_loader, val_loader, fabric, scheduler): 17 | 18 | for epoch in range(num_epochs): 19 | train_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(fabric.device) 20 | 21 | model.train() 22 | for batch_idx, (features, targets) in enumerate(train_loader): 23 | model.train() 24 | 25 | ### FORWARD AND BACK PROP 26 | logits = model(features) 27 | loss = F.cross_entropy(logits, targets) 28 | fabric.backward(loss) 29 | 30 | ### UPDATE MODEL PARAMETERS 31 | optimizer.step() 32 | optimizer.zero_grad() 33 | 34 | ### LOGGING 35 | if not batch_idx % 300: 36 | fabric.print(f"Epoch: {epoch+1:04d}/{num_epochs:04d} | Batch {batch_idx:04d}/{len(train_loader):04d} | Loss: {loss:.4f}") 37 | 38 | model.eval() 39 | with torch.no_grad(): 40 | predicted_labels = torch.argmax(logits, 1) 41 | train_acc.update(predicted_labels, targets) 42 | scheduler.step() 43 | 44 | ### MORE LOGGING 45 | model.eval() 46 | with torch.no_grad(): 47 | val_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(fabric.device) 48 | 49 | for (features, targets) in val_loader: 50 | outputs = model(features) 51 | predicted_labels = torch.argmax(outputs, 1) 52 | val_acc.update(predicted_labels, targets) 53 | 54 | fabric.print(f"Epoch: {epoch+1:04d}/{num_epochs:04d} | Train acc.: {train_acc.compute()*100:.2f}% | Val acc.: {val_acc.compute()*100:.2f}%") 55 | train_acc.reset(), val_acc.reset() 56 | 57 | 58 | if __name__ == "__main__": 59 | 60 | print("PyTorch:", torch.__version__) 61 | print("Lightning:", L.__version__) 62 | torch.set_float32_matmul_precision("medium") 63 | 64 | L.seed_everything(123) 65 | 66 | ########################## 67 | ### 1 Loading the Dataset 68 | ########################## 69 | train_transforms = transforms.Compose([transforms.Resize((224, 224)), 70 | #transforms.RandomCrop((224, 224)), 71 | transforms.ToTensor()]) 72 | 73 | test_transforms = transforms.Compose([transforms.Resize((224, 224)), 74 | #transforms.CenterCrop((224, 224)), 75 | transforms.ToTensor()]) 76 | 77 | train_loader, val_loader, test_loader = get_dataloaders_cifar10( 78 | batch_size=16, 79 | num_workers=4, 80 | train_transforms=train_transforms, 81 | test_transforms=test_transforms, 82 | validation_fraction=0.1, 83 | download=True 84 | ) 85 | 86 | ######################################### 87 | ### 2 Initializing the Model 88 | ######################################### 89 | 90 | model = vit_b_16(weights=ViT_B_16_Weights.IMAGENET1K_V1) 91 | 92 | # replace output layer 93 | model.heads.head = torch.nn.Linear(in_features=768, out_features=10) 94 | 95 | optimizer = torch.optim.Adam(model.parameters(), lr=5e-5) 96 | scheduler = ExponentialLR(optimizer, gamma=0.9) 97 | 98 | ######################################### 99 | ### 3 Launch Fabric 100 | ######################################### 101 | 102 | fabric = Fabric(accelerator="cuda", devices=1, precision="16-mixed") 103 | fabric.launch() 104 | 105 | train_loader, val_loader, test_loader = fabric.setup_dataloaders( 106 | train_loader, val_loader, test_loader) 107 | model, optimizer = fabric.setup(model, optimizer) 108 | 109 | ######################################### 110 | ### 4 Finetuning 111 | ######################################### 112 | 113 | start = time.time() 114 | train( 115 | num_epochs=3, 116 | model=model, 117 | optimizer=optimizer, 118 | train_loader=train_loader, 119 | val_loader=val_loader, 120 | fabric=fabric, 121 | scheduler=scheduler 122 | ) 123 | 124 | end = time.time() 125 | elapsed = end-start 126 | fabric.print(f"Time elapsed {elapsed/60:.2f} min") 127 | fabric.print(f"Memory used: {torch.cuda.max_memory_reserved() / 1e9:.02f} GB") 128 | 129 | ######################################### 130 | ### 5 Evaluation 131 | ######################################### 132 | 133 | with torch.no_grad(): 134 | model.eval() 135 | test_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(fabric.device) 136 | 137 | for (features, targets) in test_loader: 138 | outputs = model(features) 139 | predicted_labels = torch.argmax(outputs, 1) 140 | test_acc.update(predicted_labels, targets) 141 | 142 | fabric.print(f"Test accuracy {test_acc.compute()*100:.2f}%") -------------------------------------------------------------------------------- /precision-benchmarks/torchvision-vit16-small/04_bf16-mixed.py: -------------------------------------------------------------------------------- 1 | import time 2 | 3 | import lightning as L 4 | from lightning import Fabric 5 | import torch 6 | import torch.nn.functional as F 7 | from torch.optim.lr_scheduler import ExponentialLR 8 | import torchmetrics 9 | from torchvision import transforms 10 | from torchvision.models import vit_b_16 11 | from torchvision.models import ViT_B_16_Weights 12 | 13 | from local_utilities import get_dataloaders_cifar10 14 | 15 | 16 | def train(num_epochs, model, optimizer, train_loader, val_loader, fabric, scheduler): 17 | 18 | for epoch in range(num_epochs): 19 | train_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(fabric.device) 20 | 21 | model.train() 22 | for batch_idx, (features, targets) in enumerate(train_loader): 23 | model.train() 24 | 25 | ### FORWARD AND BACK PROP 26 | logits = model(features) 27 | loss = F.cross_entropy(logits, targets) 28 | fabric.backward(loss) 29 | 30 | ### UPDATE MODEL PARAMETERS 31 | optimizer.step() 32 | optimizer.zero_grad() 33 | 34 | ### LOGGING 35 | if not batch_idx % 300: 36 | fabric.print(f"Epoch: {epoch+1:04d}/{num_epochs:04d} | Batch {batch_idx:04d}/{len(train_loader):04d} | Loss: {loss:.4f}") 37 | 38 | model.eval() 39 | with torch.no_grad(): 40 | predicted_labels = torch.argmax(logits, 1) 41 | train_acc.update(predicted_labels, targets) 42 | scheduler.step() 43 | 44 | ### MORE LOGGING 45 | model.eval() 46 | with torch.no_grad(): 47 | val_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(fabric.device) 48 | 49 | for (features, targets) in val_loader: 50 | outputs = model(features) 51 | predicted_labels = torch.argmax(outputs, 1) 52 | val_acc.update(predicted_labels, targets) 53 | 54 | fabric.print(f"Epoch: {epoch+1:04d}/{num_epochs:04d} | Train acc.: {train_acc.compute()*100:.2f}% | Val acc.: {val_acc.compute()*100:.2f}%") 55 | train_acc.reset(), val_acc.reset() 56 | 57 | 58 | if __name__ == "__main__": 59 | 60 | print("PyTorch:", torch.__version__) 61 | print("Lightning:", L.__version__) 62 | torch.set_float32_matmul_precision("medium") 63 | 64 | L.seed_everything(123) 65 | 66 | ########################## 67 | ### 1 Loading the Dataset 68 | ########################## 69 | train_transforms = transforms.Compose([transforms.Resize((224, 224)), 70 | #transforms.RandomCrop((224, 224)), 71 | transforms.ToTensor()]) 72 | 73 | test_transforms = transforms.Compose([transforms.Resize((224, 224)), 74 | #transforms.CenterCrop((224, 224)), 75 | transforms.ToTensor()]) 76 | 77 | train_loader, val_loader, test_loader = get_dataloaders_cifar10( 78 | batch_size=16, 79 | num_workers=4, 80 | train_transforms=train_transforms, 81 | test_transforms=test_transforms, 82 | validation_fraction=0.1, 83 | download=True 84 | ) 85 | 86 | ######################################### 87 | ### 2 Initializing the Model 88 | ######################################### 89 | 90 | model = vit_b_16(weights=ViT_B_16_Weights.IMAGENET1K_V1) 91 | 92 | # replace output layer 93 | model.heads.head = torch.nn.Linear(in_features=768, out_features=10) 94 | 95 | optimizer = torch.optim.Adam(model.parameters(), lr=5e-5) 96 | scheduler = ExponentialLR(optimizer, gamma=0.9) 97 | 98 | ######################################### 99 | ### 3 Launch Fabric 100 | ######################################### 101 | 102 | fabric = Fabric(accelerator="cuda", devices=1, precision="bf16-mixed") 103 | fabric.launch() 104 | 105 | train_loader, val_loader, test_loader = fabric.setup_dataloaders( 106 | train_loader, val_loader, test_loader) 107 | model, optimizer = fabric.setup(model, optimizer) 108 | 109 | ######################################### 110 | ### 4 Finetuning 111 | ######################################### 112 | 113 | start = time.time() 114 | train( 115 | num_epochs=3, 116 | model=model, 117 | optimizer=optimizer, 118 | train_loader=train_loader, 119 | val_loader=val_loader, 120 | fabric=fabric, 121 | scheduler=scheduler 122 | ) 123 | 124 | end = time.time() 125 | elapsed = end-start 126 | fabric.print(f"Time elapsed {elapsed/60:.2f} min") 127 | fabric.print(f"Memory used: {torch.cuda.max_memory_reserved() / 1e9:.02f} GB") 128 | 129 | ######################################### 130 | ### 5 Evaluation 131 | ######################################### 132 | 133 | with torch.no_grad(): 134 | model.eval() 135 | test_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(fabric.device) 136 | 137 | for (features, targets) in test_loader: 138 | outputs = model(features) 139 | predicted_labels = torch.argmax(outputs, 1) 140 | test_acc.update(predicted_labels, targets) 141 | 142 | fabric.print(f"Test accuracy {test_acc.compute()*100:.2f}%") -------------------------------------------------------------------------------- /precision-benchmarks/torchvision-vit16-small/05_fp16-full.py: -------------------------------------------------------------------------------- 1 | import time 2 | 3 | import lightning as L 4 | from lightning import Fabric 5 | import torch 6 | import torch.nn.functional as F 7 | from torch.optim.lr_scheduler import ExponentialLR 8 | import torchmetrics 9 | from torchvision import transforms 10 | from torchvision.models import vit_b_16 11 | from torchvision.models import ViT_B_16_Weights 12 | 13 | from local_utilities import get_dataloaders_cifar10 14 | 15 | 16 | def train(num_epochs, model, optimizer, train_loader, val_loader, fabric, scheduler): 17 | 18 | for epoch in range(num_epochs): 19 | train_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(fabric.device) 20 | 21 | model.train() 22 | for batch_idx, (features, targets) in enumerate(train_loader): 23 | model.train() 24 | 25 | ### FORWARD AND BACK PROP 26 | logits = model(features) 27 | loss = F.cross_entropy(logits, targets) 28 | fabric.backward(loss) 29 | 30 | ### UPDATE MODEL PARAMETERS 31 | optimizer.step() 32 | optimizer.zero_grad() 33 | 34 | ### LOGGING 35 | if not batch_idx % 300: 36 | fabric.print(f"Epoch: {epoch+1:04d}/{num_epochs:04d} | Batch {batch_idx:04d}/{len(train_loader):04d} | Loss: {loss:.4f}") 37 | 38 | model.eval() 39 | with torch.no_grad(): 40 | predicted_labels = torch.argmax(logits, 1) 41 | train_acc.update(predicted_labels, targets) 42 | scheduler.step() 43 | 44 | ### MORE LOGGING 45 | model.eval() 46 | with torch.no_grad(): 47 | val_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(fabric.device) 48 | 49 | for (features, targets) in val_loader: 50 | outputs = model(features) 51 | predicted_labels = torch.argmax(outputs, 1) 52 | val_acc.update(predicted_labels, targets) 53 | 54 | fabric.print(f"Epoch: {epoch+1:04d}/{num_epochs:04d} | Train acc.: {train_acc.compute()*100:.2f}% | Val acc.: {val_acc.compute()*100:.2f}%") 55 | train_acc.reset(), val_acc.reset() 56 | 57 | 58 | if __name__ == "__main__": 59 | 60 | print("PyTorch:", torch.__version__) 61 | print("Lightning:", L.__version__) 62 | torch.set_float32_matmul_precision("medium") 63 | 64 | L.seed_everything(123) 65 | 66 | ########################## 67 | ### 1 Loading the Dataset 68 | ########################## 69 | train_transforms = transforms.Compose([transforms.Resize((224, 224)), 70 | #transforms.RandomCrop((224, 224)), 71 | transforms.ToTensor()]) 72 | 73 | test_transforms = transforms.Compose([transforms.Resize((224, 224)), 74 | #transforms.CenterCrop((224, 224)), 75 | transforms.ToTensor()]) 76 | 77 | train_loader, val_loader, test_loader = get_dataloaders_cifar10( 78 | batch_size=16, 79 | num_workers=4, 80 | train_transforms=train_transforms, 81 | test_transforms=test_transforms, 82 | validation_fraction=0.1, 83 | download=True 84 | ) 85 | 86 | ######################################### 87 | ### 2 Initializing the Model 88 | ######################################### 89 | 90 | model = vit_b_16(weights=ViT_B_16_Weights.IMAGENET1K_V1) 91 | 92 | # replace output layer 93 | model.heads.head = torch.nn.Linear(in_features=768, out_features=10) 94 | 95 | optimizer = torch.optim.Adam(model.parameters(), lr=5e-5) 96 | scheduler = ExponentialLR(optimizer, gamma=0.9) 97 | 98 | ######################################### 99 | ### 3 Launch Fabric 100 | ######################################### 101 | 102 | fabric = Fabric(accelerator="cuda", devices=1, precision="16-true") 103 | fabric.launch() 104 | 105 | train_loader, val_loader, test_loader = fabric.setup_dataloaders( 106 | train_loader, val_loader, test_loader) 107 | model, optimizer = fabric.setup(model, optimizer) 108 | 109 | ######################################### 110 | ### 4 Finetuning 111 | ######################################### 112 | 113 | start = time.time() 114 | train( 115 | num_epochs=3, 116 | model=model, 117 | optimizer=optimizer, 118 | train_loader=train_loader, 119 | val_loader=val_loader, 120 | fabric=fabric, 121 | scheduler=scheduler 122 | ) 123 | 124 | end = time.time() 125 | elapsed = end-start 126 | fabric.print(f"Time elapsed {elapsed/60:.2f} min") 127 | fabric.print(f"Memory used: {torch.cuda.max_memory_reserved() / 1e9:.02f} GB") 128 | 129 | ######################################### 130 | ### 5 Evaluation 131 | ######################################### 132 | 133 | with torch.no_grad(): 134 | model.eval() 135 | test_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(fabric.device) 136 | 137 | for (features, targets) in test_loader: 138 | outputs = model(features) 139 | predicted_labels = torch.argmax(outputs, 1) 140 | test_acc.update(predicted_labels, targets) 141 | 142 | fabric.print(f"Test accuracy {test_acc.compute()*100:.2f}%") -------------------------------------------------------------------------------- /precision-benchmarks/torchvision-vit16-small/06_bf16-full.py: -------------------------------------------------------------------------------- 1 | import time 2 | 3 | import lightning as L 4 | from lightning import Fabric 5 | import torch 6 | import torch.nn.functional as F 7 | from torch.optim.lr_scheduler import ExponentialLR 8 | import torchmetrics 9 | from torchvision import transforms 10 | from torchvision.models import vit_b_16 11 | from torchvision.models import ViT_B_16_Weights 12 | 13 | from local_utilities import get_dataloaders_cifar10 14 | 15 | 16 | def train(num_epochs, model, optimizer, train_loader, val_loader, fabric, scheduler): 17 | 18 | for epoch in range(num_epochs): 19 | train_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(fabric.device) 20 | 21 | model.train() 22 | for batch_idx, (features, targets) in enumerate(train_loader): 23 | model.train() 24 | 25 | ### FORWARD AND BACK PROP 26 | logits = model(features) 27 | loss = F.cross_entropy(logits, targets) 28 | fabric.backward(loss) 29 | 30 | ### UPDATE MODEL PARAMETERS 31 | optimizer.step() 32 | optimizer.zero_grad() 33 | 34 | ### LOGGING 35 | if not batch_idx % 300: 36 | fabric.print(f"Epoch: {epoch+1:04d}/{num_epochs:04d} | Batch {batch_idx:04d}/{len(train_loader):04d} | Loss: {loss:.4f}") 37 | 38 | model.eval() 39 | with torch.no_grad(): 40 | predicted_labels = torch.argmax(logits, 1) 41 | train_acc.update(predicted_labels, targets) 42 | scheduler.step() 43 | 44 | ### MORE LOGGING 45 | model.eval() 46 | with torch.no_grad(): 47 | val_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(fabric.device) 48 | 49 | for (features, targets) in val_loader: 50 | outputs = model(features) 51 | predicted_labels = torch.argmax(outputs, 1) 52 | val_acc.update(predicted_labels, targets) 53 | 54 | fabric.print(f"Epoch: {epoch+1:04d}/{num_epochs:04d} | Train acc.: {train_acc.compute()*100:.2f}% | Val acc.: {val_acc.compute()*100:.2f}%") 55 | train_acc.reset(), val_acc.reset() 56 | 57 | 58 | if __name__ == "__main__": 59 | 60 | print("PyTorch:", torch.__version__) 61 | print("Lightning:", L.__version__) 62 | torch.set_float32_matmul_precision("medium") 63 | 64 | L.seed_everything(123) 65 | 66 | ########################## 67 | ### 1 Loading the Dataset 68 | ########################## 69 | train_transforms = transforms.Compose([transforms.Resize((224, 224)), 70 | #transforms.RandomCrop((224, 224)), 71 | transforms.ToTensor()]) 72 | 73 | test_transforms = transforms.Compose([transforms.Resize((224, 224)), 74 | #transforms.CenterCrop((224, 224)), 75 | transforms.ToTensor()]) 76 | 77 | train_loader, val_loader, test_loader = get_dataloaders_cifar10( 78 | batch_size=16, 79 | num_workers=4, 80 | train_transforms=train_transforms, 81 | test_transforms=test_transforms, 82 | validation_fraction=0.1, 83 | download=True 84 | ) 85 | 86 | ######################################### 87 | ### 2 Initializing the Model 88 | ######################################### 89 | 90 | model = vit_b_16(weights=ViT_B_16_Weights.IMAGENET1K_V1) 91 | 92 | # replace output layer 93 | model.heads.head = torch.nn.Linear(in_features=768, out_features=10) 94 | 95 | optimizer = torch.optim.Adam(model.parameters(), lr=5e-5) 96 | scheduler = ExponentialLR(optimizer, gamma=0.9) 97 | 98 | ######################################### 99 | ### 3 Launch Fabric 100 | ######################################### 101 | 102 | fabric = Fabric(accelerator="cuda", devices=1, precision="bf16-true") 103 | fabric.launch() 104 | 105 | train_loader, val_loader, test_loader = fabric.setup_dataloaders( 106 | train_loader, val_loader, test_loader) 107 | model, optimizer = fabric.setup(model, optimizer) 108 | 109 | ######################################### 110 | ### 4 Finetuning 111 | ######################################### 112 | 113 | start = time.time() 114 | train( 115 | num_epochs=3, 116 | model=model, 117 | optimizer=optimizer, 118 | train_loader=train_loader, 119 | val_loader=val_loader, 120 | fabric=fabric, 121 | scheduler=scheduler 122 | ) 123 | 124 | end = time.time() 125 | elapsed = end-start 126 | fabric.print(f"Time elapsed {elapsed/60:.2f} min") 127 | fabric.print(f"Memory used: {torch.cuda.max_memory_reserved() / 1e9:.02f} GB") 128 | 129 | ######################################### 130 | ### 5 Evaluation 131 | ######################################### 132 | 133 | with torch.no_grad(): 134 | model.eval() 135 | test_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(fabric.device) 136 | 137 | for (features, targets) in test_loader: 138 | outputs = model(features) 139 | predicted_labels = torch.argmax(outputs, 1) 140 | test_acc.update(predicted_labels, targets) 141 | 142 | fabric.print(f"Test accuracy {test_acc.compute()*100:.2f}%") -------------------------------------------------------------------------------- /precision-benchmarks/torchvision-vit16-large/04_bf16-mixed.py: -------------------------------------------------------------------------------- 1 | import time 2 | 3 | import lightning as L 4 | from lightning import Fabric 5 | import torch 6 | import torch.nn.functional as F 7 | from torch.optim.lr_scheduler import ExponentialLR 8 | import torchmetrics 9 | from torchvision import transforms 10 | from torchvision.models import vit_l_16 11 | from torchvision.models import ViT_L_16_Weights 12 | 13 | from local_utilities import get_dataloaders_cifar10 14 | 15 | 16 | def train(num_epochs, model, optimizer, train_loader, val_loader, fabric, scheduler): 17 | 18 | for epoch in range(num_epochs): 19 | train_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(fabric.device) 20 | 21 | model.train() 22 | for batch_idx, (features, targets) in enumerate(train_loader): 23 | model.train() 24 | 25 | ### FORWARD AND BACK PROP 26 | logits = model(features) 27 | loss = F.cross_entropy(logits, targets) 28 | fabric.backward(loss) 29 | 30 | ### UPDATE MODEL PARAMETERS 31 | optimizer.step() 32 | optimizer.zero_grad() 33 | 34 | ### LOGGING 35 | if not batch_idx % 300: 36 | fabric.print(f"Epoch: {epoch+1:04d}/{num_epochs:04d} | Batch {batch_idx:04d}/{len(train_loader):04d} | Loss: {loss:.4f}") 37 | 38 | model.eval() 39 | with torch.no_grad(): 40 | predicted_labels = torch.argmax(logits, 1) 41 | train_acc.update(predicted_labels, targets) 42 | scheduler.step() 43 | 44 | ### MORE LOGGING 45 | model.eval() 46 | with torch.no_grad(): 47 | val_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(fabric.device) 48 | 49 | for (features, targets) in val_loader: 50 | outputs = model(features) 51 | predicted_labels = torch.argmax(outputs, 1) 52 | val_acc.update(predicted_labels, targets) 53 | 54 | fabric.print(f"Epoch: {epoch+1:04d}/{num_epochs:04d} | Train acc.: {train_acc.compute()*100:.2f}% | Val acc.: {val_acc.compute()*100:.2f}%") 55 | train_acc.reset(), val_acc.reset() 56 | 57 | 58 | if __name__ == "__main__": 59 | 60 | print("PyTorch:", torch.__version__) 61 | print("Lightning:", L.__version__) 62 | torch.set_float32_matmul_precision("medium") 63 | 64 | L.seed_everything(123) 65 | 66 | ########################## 67 | ### 1 Loading the Dataset 68 | ########################## 69 | train_transforms = transforms.Compose([transforms.Resize((224, 224)), 70 | #transforms.RandomCrop((224, 224)), 71 | transforms.ToTensor()]) 72 | 73 | test_transforms = transforms.Compose([transforms.Resize((224, 224)), 74 | #transforms.CenterCrop((224, 224)), 75 | transforms.ToTensor()]) 76 | 77 | train_loader, val_loader, test_loader = get_dataloaders_cifar10( 78 | batch_size=32, 79 | num_workers=4, 80 | train_transforms=train_transforms, 81 | test_transforms=test_transforms, 82 | validation_fraction=0.1, 83 | download=True 84 | ) 85 | 86 | ######################################### 87 | ### 2 Initializing the Model 88 | ######################################### 89 | 90 | model = vit_l_16(weights=ViT_L_16_Weights.IMAGENET1K_V1) 91 | 92 | # replace output layer 93 | model.heads.head = torch.nn.Linear(in_features=1024, out_features=10) 94 | 95 | optimizer = torch.optim.Adam(model.parameters(), lr=5e-5) 96 | scheduler = ExponentialLR(optimizer, gamma=0.9) 97 | 98 | ######################################### 99 | ### 3 Launch Fabric 100 | ######################################### 101 | 102 | fabric = Fabric(accelerator="cuda", devices=1, precision="bf16-mixed") 103 | fabric.launch() 104 | 105 | train_loader, val_loader, test_loader = fabric.setup_dataloaders( 106 | train_loader, val_loader, test_loader) 107 | model, optimizer = fabric.setup(model, optimizer) 108 | 109 | ######################################### 110 | ### 4 Finetuning 111 | ######################################### 112 | 113 | start = time.time() 114 | train( 115 | num_epochs=3, 116 | model=model, 117 | optimizer=optimizer, 118 | train_loader=train_loader, 119 | val_loader=val_loader, 120 | fabric=fabric, 121 | scheduler=scheduler 122 | ) 123 | 124 | end = time.time() 125 | elapsed = end-start 126 | fabric.print(f"Time elapsed {elapsed/60:.2f} min") 127 | fabric.print(f"Memory used: {torch.cuda.max_memory_reserved() / 1e9:.02f} GB") 128 | 129 | ######################################### 130 | ### 5 Evaluation 131 | ######################################### 132 | 133 | with torch.no_grad(): 134 | model.eval() 135 | test_acc = torchmetrics.Accuracy(task="multiclass", num_classes=10).to(fabric.device) 136 | 137 | for (features, targets) in test_loader: 138 | outputs = model(features) 139 | predicted_labels = torch.argmax(outputs, 1) 140 | test_acc.update(predicted_labels, targets) 141 | 142 | fabric.print(f"Test accuracy {test_acc.compute()*100:.2f}%") -------------------------------------------------------------------------------- /LICENSE: -------------------------------------------------------------------------------- 1 | Apache License 2 | Version 2.0, January 2004 3 | http://www.apache.org/licenses/ 4 | 5 | TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION 6 | 7 | 1. Definitions. 8 | 9 | "License" shall mean the terms and conditions for use, reproduction, 10 | and distribution as defined by Sections 1 through 9 of this document. 11 | 12 | "Licensor" shall mean the copyright owner or entity authorized by 13 | the copyright owner that is granting the License. 14 | 15 | "Legal Entity" shall mean the union of the acting entity and all 16 | other entities that control, are controlled by, or are under common 17 | control with that entity. For the purposes of this definition, 18 | "control" means (i) the power, direct or indirect, to cause the 19 | direction or management of such entity, whether by contract or 20 | otherwise, or (ii) ownership of fifty percent (50%) or more of the 21 | outstanding shares, or (iii) beneficial ownership of such entity. 22 | 23 | "You" (or "Your") shall mean an individual or Legal Entity 24 | exercising permissions granted by this License. 25 | 26 | "Source" form shall mean the preferred form for making modifications, 27 | including but not limited to software source code, documentation 28 | source, and configuration files. 29 | 30 | "Object" form shall mean any form resulting from mechanical 31 | transformation or translation of a Source form, including but 32 | not limited to compiled object code, generated documentation, 33 | and conversions to other media types. 34 | 35 | "Work" shall mean the work of authorship, whether in Source or 36 | Object form, made available under the License, as indicated by a 37 | copyright notice that is included in or attached to the work 38 | (an example is provided in the Appendix below). 39 | 40 | "Derivative Works" shall mean any work, whether in Source or Object 41 | form, that is based on (or derived from) the Work and for which the 42 | editorial revisions, annotations, elaborations, or other modifications 43 | represent, as a whole, an original work of authorship. For the purposes 44 | of this License, Derivative Works shall not include works that remain 45 | separable from, or merely link (or bind by name) to the interfaces of, 46 | the Work and Derivative Works thereof. 47 | 48 | "Contribution" shall mean any work of authorship, including 49 | the original version of the Work and any modifications or additions 50 | to that Work or Derivative Works thereof, that is intentionally 51 | submitted to Licensor for inclusion in the Work by the copyright owner 52 | or by an individual or Legal Entity authorized to submit on behalf of 53 | the copyright owner. For the purposes of this definition, "submitted" 54 | means any form of electronic, verbal, or written communication sent 55 | to the Licensor or its representatives, including but not limited to 56 | communication on electronic mailing lists, source code control systems, 57 | and issue tracking systems that are managed by, or on behalf of, the 58 | Licensor for the purpose of discussing and improving the Work, but 59 | excluding communication that is conspicuously marked or otherwise 60 | designated in writing by the copyright owner as "Not a Contribution." 61 | 62 | "Contributor" shall mean Licensor and any individual or Legal Entity 63 | on behalf of whom a Contribution has been received by Licensor and 64 | subsequently incorporated within the Work. 65 | 66 | 2. Grant of Copyright License. Subject to the terms and conditions of 67 | this License, each Contributor hereby grants to You a perpetual, 68 | worldwide, non-exclusive, no-charge, royalty-free, irrevocable 69 | copyright license to reproduce, prepare Derivative Works of, 70 | publicly display, publicly perform, sublicense, and distribute the 71 | Work and such Derivative Works in Source or Object form. 72 | 73 | 3. Grant of Patent License. Subject to the terms and conditions of 74 | this License, each Contributor hereby grants to You a perpetual, 75 | worldwide, non-exclusive, no-charge, royalty-free, irrevocable 76 | (except as stated in this section) patent license to make, have made, 77 | use, offer to sell, sell, import, and otherwise transfer the Work, 78 | where such license applies only to those patent claims licensable 79 | by such Contributor that are necessarily infringed by their 80 | Contribution(s) alone or by combination of their Contribution(s) 81 | with the Work to which such Contribution(s) was submitted. If You 82 | institute patent litigation against any entity (including a 83 | cross-claim or counterclaim in a lawsuit) alleging that the Work 84 | or a Contribution incorporated within the Work constitutes direct 85 | or contributory patent infringement, then any patent licenses 86 | granted to You under this License for that Work shall terminate 87 | as of the date such litigation is filed. 88 | 89 | 4. Redistribution. You may reproduce and distribute copies of the 90 | Work or Derivative Works thereof in any medium, with or without 91 | modifications, and in Source or Object form, provided that You 92 | meet the following conditions: 93 | 94 | (a) You must give any other recipients of the Work or 95 | Derivative Works a copy of this License; and 96 | 97 | (b) You must cause any modified files to carry prominent notices 98 | stating that You changed the files; and 99 | 100 | (c) You must retain, in the Source form of any Derivative Works 101 | that You distribute, all copyright, patent, trademark, and 102 | attribution notices from the Source form of the Work, 103 | excluding those notices that do not pertain to any part of 104 | the Derivative Works; and 105 | 106 | (d) If the Work includes a "NOTICE" text file as part of its 107 | distribution, then any Derivative Works that You distribute must 108 | include a readable copy of the attribution notices contained 109 | within such NOTICE file, excluding those notices that do not 110 | pertain to any part of the Derivative Works, in at least one 111 | of the following places: within a NOTICE text file distributed 112 | as part of the Derivative Works; within the Source form or 113 | documentation, if provided along with the Derivative Works; or, 114 | within a display generated by the Derivative Works, if and 115 | wherever such third-party notices normally appear. The contents 116 | of the NOTICE file are for informational purposes only and 117 | do not modify the License. You may add Your own attribution 118 | notices within Derivative Works that You distribute, alongside 119 | or as an addendum to the NOTICE text from the Work, provided 120 | that such additional attribution notices cannot be construed 121 | as modifying the License. 122 | 123 | You may add Your own copyright statement to Your modifications and 124 | may provide additional or different license terms and conditions 125 | for use, reproduction, or distribution of Your modifications, or 126 | for any such Derivative Works as a whole, provided Your use, 127 | reproduction, and distribution of the Work otherwise complies with 128 | the conditions stated in this License. 129 | 130 | 5. Submission of Contributions. Unless You explicitly state otherwise, 131 | any Contribution intentionally submitted for inclusion in the Work 132 | by You to the Licensor shall be under the terms and conditions of 133 | this License, without any additional terms or conditions. 134 | Notwithstanding the above, nothing herein shall supersede or modify 135 | the terms of any separate license agreement you may have executed 136 | with Licensor regarding such Contributions. 137 | 138 | 6. Trademarks. This License does not grant permission to use the trade 139 | names, trademarks, service marks, or product names of the Licensor, 140 | except as required for reasonable and customary use in describing the 141 | origin of the Work and reproducing the content of the NOTICE file. 142 | 143 | 7. Disclaimer of Warranty. Unless required by applicable law or 144 | agreed to in writing, Licensor provides the Work (and each 145 | Contributor provides its Contributions) on an "AS IS" BASIS, 146 | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or 147 | implied, including, without limitation, any warranties or conditions 148 | of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A 149 | PARTICULAR PURPOSE. You are solely responsible for determining the 150 | appropriateness of using or redistributing the Work and assume any 151 | risks associated with Your exercise of permissions under this License. 152 | 153 | 8. Limitation of Liability. In no event and under no legal theory, 154 | whether in tort (including negligence), contract, or otherwise, 155 | unless required by applicable law (such as deliberate and grossly 156 | negligent acts) or agreed to in writing, shall any Contributor be 157 | liable to You for damages, including any direct, indirect, special, 158 | incidental, or consequential damages of any character arising as a 159 | result of this License or out of the use or inability to use the 160 | Work (including but not limited to damages for loss of goodwill, 161 | work stoppage, computer failure or malfunction, or any and all 162 | other commercial damages or losses), even if such Contributor 163 | has been advised of the possibility of such damages. 164 | 165 | 9. Accepting Warranty or Additional Liability. While redistributing 166 | the Work or Derivative Works thereof, You may choose to offer, 167 | and charge a fee for, acceptance of support, warranty, indemnity, 168 | or other liability obligations and/or rights consistent with this 169 | License. However, in accepting such obligations, You may act only 170 | on Your own behalf and on Your sole responsibility, not on behalf 171 | of any other Contributor, and only if You agree to indemnify, 172 | defend, and hold each Contributor harmless for any liability 173 | incurred by, or claims asserted against, such Contributor by reason 174 | of your accepting any such warranty or additional liability. 175 | 176 | END OF TERMS AND CONDITIONS 177 | 178 | APPENDIX: How to apply the Apache License to your work. 179 | 180 | To apply the Apache License to your work, attach the following 181 | boilerplate notice, with the fields enclosed by brackets "[]" 182 | replaced with your own identifying information. (Don't include 183 | the brackets!) The text should be enclosed in the appropriate 184 | comment syntax for the file format. We also recommend that a 185 | file or class name and description of purpose be included on the 186 | same "printed page" as the copyright notice for easier 187 | identification within third-party archives. 188 | 189 | Copyright [yyyy] [name of copyright owner] 190 | 191 | Licensed under the Apache License, Version 2.0 (the "License"); 192 | you may not use this file except in compliance with the License. 193 | You may obtain a copy of the License at 194 | 195 | http://www.apache.org/licenses/LICENSE-2.0 196 | 197 | Unless required by applicable law or agreed to in writing, software 198 | distributed under the License is distributed on an "AS IS" BASIS, 199 | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 200 | See the License for the specific language governing permissions and 201 | limitations under the License. 202 | --------------------------------------------------------------------------------