[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-2122":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":16,"subscribersCount":16,"size":16,"stars1d":16,"stars7d":17,"stars30d":18,"stars90d":16,"forks30d":16,"starsTrendScore":19,"compositeScore":20,"rankGlobal":10,"rankLanguage":10,"license":21,"archived":22,"fork":22,"defaultBranch":23,"hasWiki":22,"hasPages":22,"topics":24,"createdAt":10,"pushedAt":10,"updatedAt":32,"readmeContent":33,"aiSummary":34,"trendingCount":16,"starSnapshotCount":16,"syncStatus":35,"lastSyncTime":36,"discoverSource":37},2122,"pytorch-lightning","Lightning-AI\u002Fpytorch-lightning","Lightning-AI","Pretrain, finetune ANY AI model of ANY size on 1 or 10,000+ GPUs with zero code changes.","https:\u002F\u002Flightning.ai\u002Fpytorch-lightning\u002F?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme",null,"Python",31180,3736,256,835,0,10,54,4,45,"Apache License 2.0",false,"master",[25,26,27,28,29,30,31],"ai","artificial-intelligence","data-science","deep-learning","machine-learning","python","pytorch","2026-06-12 02:00:37","\u003Cdiv align=\"center\">\n\n\u003Cimg alt=\"Lightning\" src=\"https:\u002F\u002Fpl-bolts-doc-images.s3.us-east-2.amazonaws.com\u002Fapp-2\u002Fptl_banner.png\" width=\"800px\" style=\"max-width: 100%;\">\n\n\u003Cbr\u002F>\n\u003Cbr\u002F>\n\n**The deep learning framework to pretrain and finetune AI models.**\n\n**Serving models?** Use [LitServe](https:\u002F\u002Fgithub.com\u002FLightning-AI\u002Flitserve?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme) to build custom inference servers in pure Python.\n\n______________________________________________________________________\n\n\u003Cp align=\"center\">\n    \u003Ca href=\"#quick-start\" style=\"margin: 0 10px;\">Quick start\u003C\u002Fa> •\n  \u003Ca href=\"#examples\">Examples\u003C\u002Fa> •\n  \u003Ca href=\"#why-pytorch-lightning\">PyTorch Lightning\u003C\u002Fa> •\n  \u003Ca href=\"#lightning-fabric-expert-control\">Fabric\u003C\u002Fa> •\n  \u003Ca href=\"https:\u002F\u002Flightning.ai\u002F?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme\">Lightning Cloud\u003C\u002Fa> •   \n  \u003Ca href=\"#community\">Community\u003C\u002Fa> •\n  \u003Ca href=\"https:\u002F\u002Fpytorch-lightning.readthedocs.io\u002Fen\u002Fstable\u002F\">Docs\u003C\u002Fa>\n\u003C\u002Fp>\n\n\u003C!-- DO NOT ADD CONDA DOWNLOADS... README CHANGES MUST BE APPROVED BY EDEN OR WILL -->\n\n[![PyPI - Python Version](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fpyversions\u002Fpytorch-lightning)](https:\u002F\u002Fpypi.org\u002Fproject\u002Fpytorch-lightning\u002F)\n[![PyPI Status](https:\u002F\u002Fbadge.fury.io\u002Fpy\u002Fpytorch-lightning.svg)](https:\u002F\u002Fbadge.fury.io\u002Fpy\u002Fpytorch-lightning)\n[![PyPI - Downloads](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fdm\u002Fpytorch-lightning)](https:\u002F\u002Fpepy.tech\u002Fproject\u002Fpytorch-lightning)\n[![Conda](https:\u002F\u002Fimg.shields.io\u002Fconda\u002Fv\u002Fconda-forge\u002Flightning?label=conda&color=success)](https:\u002F\u002Fanaconda.org\u002Fconda-forge\u002Flightning)\n[![codecov](https:\u002F\u002Fcodecov.io\u002Fgh\u002FLightning-AI\u002Fpytorch-lightning\u002Fgraph\u002Fbadge.svg?token=SmzX8mnKlA)](https:\u002F\u002Fcodecov.io\u002Fgh\u002FLightning-AI\u002Fpytorch-lightning)\n\n[![Discord](https:\u002F\u002Fimg.shields.io\u002Fdiscord\u002F1077906959069626439?style=plastic)](https:\u002F\u002Fdiscord.gg\u002FVptPCZkGNa)\n![GitHub commit activity](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fcommit-activity\u002Fw\u002Flightning-ai\u002Flightning)\n[![license](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-Apache%202.0-blue.svg)](https:\u002F\u002Fgithub.com\u002FLightning-AI\u002Fpytorch-lightning\u002Fblob\u002Fmaster\u002FLICENSE)\n\n\u003C!--\n[![CodeFactor](https:\u002F\u002Fwww.codefactor.io\u002Frepository\u002Fgithub\u002FLightning-AI\u002Flightning\u002Fbadge)](https:\u002F\u002Fwww.codefactor.io\u002Frepository\u002Fgithub\u002FLightning-AI\u002Flightning)\n-->\n\n\u003C\u002Fdiv>\n\n\u003Cdiv align=\"center\">\n  \n\u003Cp align=\"center\">\n\n \n\n\u003Ca target=\"_blank\" href=\"https:\u002F\u002Flightning.ai\u002Fdocs\u002Fpytorch\u002Flatest\u002Fstarter\u002Fintroduction.html?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme#define-a-lightningmodule\">\n  \u003Cimg src=\"https:\u002F\u002Fpl-bolts-doc-images.s3.us-east-2.amazonaws.com\u002Fapp-2\u002Fget-started-badge.svg\" height=\"36px\" alt=\"Get started\"\u002F>\n\u003C\u002Fa>\n\n\u003C\u002Fp>\n\n\u003C\u002Fdiv>\n\n \n\n\u003Ca id=\"why-pytorch-lightning\">\u003C\u002Fa>\n# Why PyTorch Lightning?   \nTraining models in plain PyTorch requires writing and maintaining a lot of repetitive engineering code. Handling backpropagation, mixed precision, multi-GPU, and distributed training is error-prone and often reimplemented for every project. PyTorch Lightning organizes PyTorch code to automate this infrastructure while keeping full control over your model logic. You write the science. Lightning handles the engineering, and scales from CPU to multi-node GPUs without changing your core code. PyTorch experts can still opt into [expert-level control](#lightning-fabric-expert-control).   \n\nFun analogy: If PyTorch is Javascript, PyTorch Lightning is ReactJS or NextJS.\n\n# Looking for GPUs?\n[Lightning Cloud](https:\u002F\u002Flightning.ai\u002F?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme) is the easiest way to run PyTorch Lightning without managing infrastructure. Start training with one command and get GPUs, autoscaling, monitoring, and a free tier. No cloud setup required.\n\nYou can also run PyTorch Lightning on your own hardware or cloud.\n\n# Lightning has 2 core packages\n\n[PyTorch Lightning: Train and deploy PyTorch at scale](#why-pytorch-lightning).\n\u003Cbr\u002F>\n[Lightning Fabric: Expert control](#lightning-fabric-expert-control).\n\nLightning gives you granular control over how much abstraction you want to add over PyTorch.\n\n\u003Cdiv align=\"center\">\n    \u003Cimg src=\"https:\u002F\u002Fpl-public-data.s3.amazonaws.com\u002Fassets_lightning\u002Fcontinuum.png\" width=\"80%\">\n\u003C\u002Fdiv>\n\n \n\n# Quick start\nInstall Lightning:\n\n```bash\npip install lightning\n```\n\n\u003C!-- following section will be skipped from PyPI description -->\n\n\u003Cdetails>\n  \u003Csummary>Advanced install options\u003C\u002Fsummary>\n    \u003C!-- following section will be skipped from PyPI description -->\n\n#### Install with optional dependencies\n\n```bash\npip install lightning['extra']\n```\n\n#### Conda\n\n```bash\nconda install lightning -c conda-forge\n```\n\n#### Install stable version\n\nInstall future release from the source\n\n```bash\npip install https:\u002F\u002Fgithub.com\u002FLightning-AI\u002Flightning\u002Farchive\u002Frefs\u002Fheads\u002Frelease\u002Fstable.zip -U\n```\n\n#### Install bleeding-edge\n\nInstall nightly from the source (no guarantees)\n\n```bash\npip install https:\u002F\u002Fgithub.com\u002FLightning-AI\u002Flightning\u002Farchive\u002Frefs\u002Fheads\u002Fmaster.zip -U\n```\n\nor from testing PyPI\n\n```bash\npip install -iU https:\u002F\u002Ftest.pypi.org\u002Fsimple\u002F pytorch-lightning\n```\n\n\u003C\u002Fdetails>\n\u003C!-- end skipping PyPI description -->\n\n### PyTorch Lightning example\nDefine the training workflow. Here's a toy example ([explore real examples](https:\u002F\u002Flightning.ai\u002Flightning-ai\u002Fstudios?view=public&section=featured&query=pytorch+lightning&utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme)):\n\n```python\n# main.py\n# ! pip install torchvision\nimport torch, torch.nn as nn, torch.utils.data as data, torchvision as tv, torch.nn.functional as F\nimport lightning as L\n\n# --------------------------------\n# Step 1: Define a LightningModule\n# --------------------------------\n# A LightningModule (nn.Module subclass) defines a full *system*\n# (ie: an LLM, diffusion model, autoencoder, or simple image classifier).\n\n\nclass LitAutoEncoder(L.LightningModule):\n    def __init__(self):\n        super().__init__()\n        self.encoder = nn.Sequential(nn.Linear(28 * 28, 128), nn.ReLU(), nn.Linear(128, 3))\n        self.decoder = nn.Sequential(nn.Linear(3, 128), nn.ReLU(), nn.Linear(128, 28 * 28))\n\n    def forward(self, x):\n        # in lightning, forward defines the prediction\u002Finference actions\n        embedding = self.encoder(x)\n        return embedding\n\n    def training_step(self, batch, batch_idx):\n        # training_step defines the train loop. It is independent of forward\n        x, _ = batch\n        x = x.view(x.size(0), -1)\n        z = self.encoder(x)\n        x_hat = self.decoder(z)\n        loss = F.mse_loss(x_hat, x)\n        self.log(\"train_loss\", loss)\n        return loss\n\n    def configure_optimizers(self):\n        optimizer = torch.optim.Adam(self.parameters(), lr=1e-3)\n        return optimizer\n\n\n# -------------------\n# Step 2: Define data\n# -------------------\ndataset = tv.datasets.MNIST(\".\", download=True, transform=tv.transforms.ToTensor())\ntrain, val = data.random_split(dataset, [55000, 5000])\n\n# -------------------\n# Step 3: Train\n# -------------------\nautoencoder = LitAutoEncoder()\ntrainer = L.Trainer()\ntrainer.fit(autoencoder, data.DataLoader(train), data.DataLoader(val))\n```\n\nRun the model on your terminal\n\n```bash\npip install torchvision\npython main.py\n```\n\n \n\n\n# Convert from PyTorch to PyTorch Lightning\n\nPyTorch Lightning is just organized PyTorch - Lightning disentangles PyTorch code to decouple the science from the engineering.\n\n![PT to PL](docs\u002Fsource-pytorch\u002F_static\u002Fimages\u002Fgeneral\u002Fpl_quick_start_full_compressed.gif)\n\n \n\n----\n\n### Examples\nExplore various types of training possible with PyTorch Lightning. Pretrain and finetune ANY kind of model to perform ANY task like classification, segmentation, summarization and more:    \n\n| Task | Description | Run |\n|------|--------------|-----|\n| [Hello world](https:\u002F\u002Flightning.ai\u002Flightning-ai\u002Fstudios\u002Fpytorch-lightning-hello-world?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme) | Pretrain - Hello world example | \u003Ca target=\"_blank\" href=\"https:\u002F\u002Flightning.ai\u002Flightning-ai\u002Fstudios\u002Fpytorch-lightning-hello-world?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme\">\u003Cimg src=\"https:\u002F\u002Fpl-bolts-doc-images.s3.us-east-2.amazonaws.com\u002Fapp-2\u002Fstudio-badge.svg\" alt=\"Open In Studio\"\u002F>\u003C\u002Fa> |\n| [Image classification](https:\u002F\u002Flightning.ai\u002Flightning-ai\u002Fstudios\u002Fimage-classification-with-pytorch-lightning?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme) | Finetune - ResNet-34 model to classify images of cars | \u003Ca target=\"_blank\" href=\"https:\u002F\u002Flightning.ai\u002Flightning-ai\u002Fstudios\u002Fimage-classification-with-pytorch-lightning?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme\">\u003Cimg src=\"https:\u002F\u002Fpl-bolts-doc-images.s3.us-east-2.amazonaws.com\u002Fapp-2\u002Fstudio-badge.svg\" alt=\"Open In Studio\"\u002F>\u003C\u002Fa> |\n| [Image segmentation](https:\u002F\u002Flightning.ai\u002Flightning-ai\u002Fstudios\u002Fimage-segmentation-with-pytorch-lightning?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme) | Finetune - ResNet-50 model to segment images | \u003Ca target=\"_blank\" href=\"https:\u002F\u002Flightning.ai\u002Flightning-ai\u002Fstudios\u002Fimage-segmentation-with-pytorch-lightning?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme\">\u003Cimg src=\"https:\u002F\u002Fpl-bolts-doc-images.s3.us-east-2.amazonaws.com\u002Fapp-2\u002Fstudio-badge.svg\" alt=\"Open In Studio\"\u002F>\u003C\u002Fa> |\n| [Object detection](https:\u002F\u002Flightning.ai\u002Flightning-ai\u002Fstudios\u002Fobject-detection-with-pytorch-lightning?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme) | Finetune - Faster R-CNN model to detect objects | \u003Ca target=\"_blank\" href=\"https:\u002F\u002Flightning.ai\u002Flightning-ai\u002Fstudios\u002Fobject-detection-with-pytorch-lightning?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme\">\u003Cimg src=\"https:\u002F\u002Fpl-bolts-doc-images.s3.us-east-2.amazonaws.com\u002Fapp-2\u002Fstudio-badge.svg\" alt=\"Open In Studio\"\u002F>\u003C\u002Fa> |\n| [Text classification](https:\u002F\u002Flightning.ai\u002Flightning-ai\u002Fstudios\u002Ftext-classification-with-pytorch-lightning?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme) | Finetune - text classifier (BERT model) | \u003Ca target=\"_blank\" href=\"https:\u002F\u002Flightning.ai\u002Flightning-ai\u002Fstudios\u002Ftext-classification-with-pytorch-lightning?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme\">\u003Cimg src=\"https:\u002F\u002Fpl-bolts-doc-images.s3.us-east-2.amazonaws.com\u002Fapp-2\u002Fstudio-badge.svg\" alt=\"Open In Studio\"\u002F>\u003C\u002Fa> |\n| [Text summarization](https:\u002F\u002Flightning.ai\u002Flightning-ai\u002Fstudios\u002Ftext-summarization-with-pytorch-lightning?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme) | Finetune - text summarization (Hugging Face transformer model) | \u003Ca target=\"_blank\" href=\"https:\u002F\u002Flightning.ai\u002Flightning-ai\u002Fstudios\u002Ftext-summarization-with-pytorch-lightning?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme\">\u003Cimg src=\"https:\u002F\u002Fpl-bolts-doc-images.s3.us-east-2.amazonaws.com\u002Fapp-2\u002Fstudio-badge.svg\" alt=\"Open In Studio\"\u002F>\u003C\u002Fa> |\n| [Audio generation](https:\u002F\u002Flightning.ai\u002Flightning-ai\u002Fstudios\u002Ffinetune-a-personal-ai-music-generator?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme) | Finetune - audio generator (transformer model) | \u003Ca target=\"_blank\" href=\"https:\u002F\u002Flightning.ai\u002Flightning-ai\u002Fstudios\u002Ffinetune-a-personal-ai-music-generator?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme\">\u003Cimg src=\"https:\u002F\u002Fpl-bolts-doc-images.s3.us-east-2.amazonaws.com\u002Fapp-2\u002Fstudio-badge.svg\" alt=\"Open In Studio\"\u002F>\u003C\u002Fa> |\n| [LLM finetuning](https:\u002F\u002Flightning.ai\u002Flightning-ai\u002Fstudios\u002Ffinetune-an-llm-with-pytorch-lightning?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme) | Finetune - LLM (Meta Llama 3.1 8B) | \u003Ca target=\"_blank\" href=\"https:\u002F\u002Flightning.ai\u002Flightning-ai\u002Fstudios\u002Ffinetune-an-llm-with-pytorch-lightning?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme\">\u003Cimg src=\"https:\u002F\u002Fpl-bolts-doc-images.s3.us-east-2.amazonaws.com\u002Fapp-2\u002Fstudio-badge.svg\" alt=\"Open In Studio\"\u002F>\u003C\u002Fa> |\n| [Image generation](https:\u002F\u002Flightning.ai\u002Flightning-ai\u002Fstudios\u002Ftrain-a-diffusion-model-with-pytorch-lightning?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme) | Pretrain - Image generator (diffusion model) | \u003Ca target=\"_blank\" href=\"https:\u002F\u002Flightning.ai\u002Flightning-ai\u002Fstudios\u002Ftrain-a-diffusion-model-with-pytorch-lightning?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme\">\u003Cimg src=\"https:\u002F\u002Fpl-bolts-doc-images.s3.us-east-2.amazonaws.com\u002Fapp-2\u002Fstudio-badge.svg\" alt=\"Open In Studio\"\u002F>\u003C\u002Fa> |\n| [Recommendation system](https:\u002F\u002Flightning.ai\u002Flightning-ai\u002Fstudios\u002Frecommendation-system-with-pytorch-lightning?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme) | Train - recommendation system (factorization and embedding) | \u003Ca target=\"_blank\" href=\"https:\u002F\u002Flightning.ai\u002Flightning-ai\u002Fstudios\u002Frecommendation-system-with-pytorch-lightning?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme\">\u003Cimg src=\"https:\u002F\u002Fpl-bolts-doc-images.s3.us-east-2.amazonaws.com\u002Fapp-2\u002Fstudio-badge.svg\" alt=\"Open In Studio\"\u002F>\u003C\u002Fa> |\n| [Time-series forecasting](https:\u002F\u002Flightning.ai\u002Flightning-ai\u002Fstudios\u002Ftime-series-forecasting-with-pytorch-lightning?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme) | Train - Time-series forecasting with LSTM | \u003Ca target=\"_blank\" href=\"https:\u002F\u002Flightning.ai\u002Flightning-ai\u002Fstudios\u002Ftime-series-forecasting-with-pytorch-lightning?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme\">\u003Cimg src=\"https:\u002F\u002Fpl-bolts-doc-images.s3.us-east-2.amazonaws.com\u002Fapp-2\u002Fstudio-badge.svg\" alt=\"Open In Studio\"\u002F>\u003C\u002Fa> |\n\n\n______________________________________________________________________\n\n## Advanced features\n\nLightning has over [40+ advanced features](https:\u002F\u002Flightning.ai\u002Fdocs\u002Fpytorch\u002Fstable\u002Fcommon\u002Ftrainer.html?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme#trainer-flags)\ndesigned for professional AI research at scale.\n\nHere are some examples:\n\n\u003Cdiv align=\"center\">\n    \u003Cimg src=\"https:\u002F\u002Fpl-bolts-doc-images.s3.us-east-2.amazonaws.com\u002Ffeatures_2.jpg\" max-height=\"600px\">\n  \u003C\u002Fdiv>\n\n\u003Cdetails>\n  \u003Csummary>Train on 1000s of GPUs without code changes\u003C\u002Fsummary>\n\n```python\n# 8 GPUs\n# no code changes needed\ntrainer = Trainer(accelerator=\"gpu\", devices=8)\n\n# 256 GPUs\ntrainer = Trainer(accelerator=\"gpu\", devices=8, num_nodes=32)\n```\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Train on other accelerators like TPUs without code changes\u003C\u002Fsummary>\n\n```python\n# no code changes needed\ntrainer = Trainer(accelerator=\"tpu\", devices=8)\n```\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>16-bit precision\u003C\u002Fsummary>\n\n```python\n# no code changes needed\ntrainer = Trainer(precision=16)\n```\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Experiment managers\u003C\u002Fsummary>\n\n```python\nfrom lightning import loggers\n\n# litlogger\ntrainer = Trainer(logger=LitLogger())\n\n# tensorboard\ntrainer = Trainer(logger=TensorBoardLogger(\"logs\u002F\"))\n\n# weights and biases\ntrainer = Trainer(logger=loggers.WandbLogger())\n\n# comet\ntrainer = Trainer(logger=loggers.CometLogger())\n\n# mlflow\ntrainer = Trainer(logger=loggers.MLFlowLogger())\n\n# ... and dozens more\n```\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\n\u003Csummary>Early Stopping\u003C\u002Fsummary>\n\n```python\nes = EarlyStopping(monitor=\"val_loss\")\ntrainer = Trainer(callbacks=[es])\n```\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Checkpointing\u003C\u002Fsummary>\n\n```python\ncheckpointing = ModelCheckpoint(monitor=\"val_loss\")\ntrainer = Trainer(callbacks=[checkpointing])\n```\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Export to torchscript (JIT) (production use)\u003C\u002Fsummary>\n\n```python\n# torchscript\nautoencoder = LitAutoEncoder()\ntorch.jit.save(autoencoder.to_torchscript(), \"model.pt\")\n```\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Export to ONNX (production use)\u003C\u002Fsummary>\n\n```python\n# onnx\nwith tempfile.NamedTemporaryFile(suffix=\".onnx\", delete=False) as tmpfile:\n    autoencoder = LitAutoEncoder()\n    input_sample = torch.randn((1, 64))\n    autoencoder.to_onnx(tmpfile.name, input_sample, export_params=True)\n    os.path.isfile(tmpfile.name)\n```\n\n\u003C\u002Fdetails>\n\n______________________________________________________________________\n\n## Advantages over unstructured PyTorch\n\n- Models become hardware agnostic\n- Code is clear to read because engineering code is abstracted away\n- Easier to reproduce\n- Make fewer mistakes because lightning handles the tricky engineering\n- Keeps all the flexibility (LightningModules are still PyTorch modules), but removes a ton of boilerplate\n- Lightning has dozens of integrations with popular machine learning tools.\n- [Tested rigorously with every new PR](https:\u002F\u002Fgithub.com\u002FLightning-AI\u002Flightning\u002Ftree\u002Fmaster\u002Ftests). We test every combination of PyTorch and Python supported versions, every OS, multi GPUs and even TPUs.\n- Minimal running speed overhead (about 300 ms per epoch compared with pure PyTorch).\n\n______________________________________________________________________\n\n\u003Cdiv align=\"center\">\n    \u003Ca href=\"https:\u002F\u002Flightning.ai\u002Fdocs\u002Fpytorch\u002Fstable\u002F?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme\">Read the PyTorch Lightning docs\u003C\u002Fa>\n\u003C\u002Fdiv>\n\n______________________________________________________________________\n\n \n \n\n# Lightning Fabric: Expert control\n\nRun on any device at any scale with expert-level control over PyTorch training loop and scaling strategy. You can even write your own Trainer.\n\nFabric is designed for the most complex models like foundation model scaling, LLMs, diffusion, transformers, reinforcement learning, active learning. Of any size.\n\n\u003Ctable>\n\u003Ctr>\n\u003Cth>What to change\u003C\u002Fth>\n\u003Cth>Resulting Fabric Code (copy me!)\u003C\u002Fth>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\n\u003Csub>\n\n```diff\n+ import lightning as L\n  import torch; import torchvision as tv\n\n dataset = tv.datasets.CIFAR10(\"data\", download=True,\n                               train=True,\n                               transform=tv.transforms.ToTensor())\n\n+ fabric = L.Fabric()\n+ fabric.launch()\n\n  model = tv.models.resnet18()\n  optimizer = torch.optim.SGD(model.parameters(), lr=0.001)\n- device = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n- model.to(device)\n+ model, optimizer = fabric.setup(model, optimizer)\n\n  dataloader = torch.utils.data.DataLoader(dataset, batch_size=8)\n+ dataloader = fabric.setup_dataloaders(dataloader)\n\n  model.train()\n  num_epochs = 10\n  for epoch in range(num_epochs):\n      for batch in dataloader:\n          inputs, labels = batch\n-         inputs, labels = inputs.to(device), labels.to(device)\n          optimizer.zero_grad()\n          outputs = model(inputs)\n          loss = torch.nn.functional.cross_entropy(outputs, labels)\n-         loss.backward()\n+         fabric.backward(loss)\n          optimizer.step()\n          print(loss.data)\n```\n\n\u003C\u002Fsub>\n\u003Ctd>\n\u003Csub>\n\n```Python\nimport lightning as L\nimport torch; import torchvision as tv\n\ndataset = tv.datasets.CIFAR10(\"data\", download=True,\n                              train=True,\n                              transform=tv.transforms.ToTensor())\n\nfabric = L.Fabric()\nfabric.launch()\n\nmodel = tv.models.resnet18()\noptimizer = torch.optim.SGD(model.parameters(), lr=0.001)\nmodel, optimizer = fabric.setup(model, optimizer)\n\ndataloader = torch.utils.data.DataLoader(dataset, batch_size=8)\ndataloader = fabric.setup_dataloaders(dataloader)\n\nmodel.train()\nnum_epochs = 10\nfor epoch in range(num_epochs):\n    for batch in dataloader:\n        inputs, labels = batch\n        optimizer.zero_grad()\n        outputs = model(inputs)\n        loss = torch.nn.functional.cross_entropy(outputs, labels)\n        fabric.backward(loss)\n        optimizer.step()\n        print(loss.data)\n```\n\n\u003C\u002Fsub>\n\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003C\u002Ftable>\n\n## Key features\n\n\u003Cdetails>\n  \u003Csummary>Easily switch from running on CPU to GPU (Apple Silicon, CUDA, …), TPU, multi-GPU or even multi-node training\u003C\u002Fsummary>\n\n```python\n# Use your available hardware\n# no code changes needed\nfabric = Fabric()\n\n# Run on GPUs (CUDA or MPS)\nfabric = Fabric(accelerator=\"gpu\")\n\n# 8 GPUs\nfabric = Fabric(accelerator=\"gpu\", devices=8)\n\n# 256 GPUs, multi-node\nfabric = Fabric(accelerator=\"gpu\", devices=8, num_nodes=32)\n\n# Run on TPUs\nfabric = Fabric(accelerator=\"tpu\")\n```\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Use state-of-the-art distributed training strategies (DDP, FSDP, DeepSpeed) and mixed precision out of the box\u003C\u002Fsummary>\n\n```python\n# Use state-of-the-art distributed training techniques\nfabric = Fabric(strategy=\"ddp\")\nfabric = Fabric(strategy=\"deepspeed\")\nfabric = Fabric(strategy=\"fsdp\")\n\n# Switch the precision\nfabric = Fabric(precision=\"16-mixed\")\nfabric = Fabric(precision=\"64\")\n```\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>All the device logic boilerplate is handled for you\u003C\u002Fsummary>\n\n```diff\n  # no more of this!\n- model.to(device)\n- batch.to(device)\n```\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Build your own custom Trainer using Fabric primitives for training checkpointing, logging, and more\u003C\u002Fsummary>\n\n```python\nimport lightning as L\n\n\nclass MyCustomTrainer:\n    def __init__(self, accelerator=\"auto\", strategy=\"auto\", devices=\"auto\", precision=\"32-true\"):\n        self.fabric = L.Fabric(accelerator=accelerator, strategy=strategy, devices=devices, precision=precision)\n\n    def fit(self, model, optimizer, dataloader, max_epochs):\n        self.fabric.launch()\n\n        model, optimizer = self.fabric.setup(model, optimizer)\n        dataloader = self.fabric.setup_dataloaders(dataloader)\n        model.train()\n\n        for epoch in range(max_epochs):\n            for batch in dataloader:\n                input, target = batch\n                optimizer.zero_grad()\n                output = model(input)\n                loss = loss_fn(output, target)\n                self.fabric.backward(loss)\n                optimizer.step()\n```\n\nYou can find a more extensive example in our [examples](examples\u002Ffabric\u002Fbuild_your_own_trainer)\n\n\u003C\u002Fdetails>\n\n______________________________________________________________________\n\n\u003Cdiv align=\"center\">\n    \u003Ca href=\"https:\u002F\u002Flightning.ai\u002Fdocs\u002Ffabric\u002Fstable\u002F?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme\">Read the Lightning Fabric docs\u003C\u002Fa>\n\u003C\u002Fdiv>\n\n______________________________________________________________________\n\n \n \n\n## Examples\n\n###### Self-supervised Learning\n\n- [CPC transforms](https:\u002F\u002Flightning-bolts.readthedocs.io\u002Fen\u002Fstable\u002Ftransforms\u002Fself_supervised.html#cpc-transforms)\n- [Moco v2 transforms](https:\u002F\u002Flightning-bolts.readthedocs.io\u002Fen\u002Fstable\u002Ftransforms\u002Fself_supervised.html#moco-v2-transforms)\n- [SimCLR transforms](https:\u002F\u002Flightning-bolts.readthedocs.io\u002Fen\u002Fstable\u002Ftransforms\u002Fself_supervised.html#simclr-transforms)\n\n###### Convolutional Architectures\n\n- [GPT-2](https:\u002F\u002Flightning-bolts.readthedocs.io\u002Fen\u002Fstable\u002Fmodels\u002Fconvolutional.html#gpt-2)\n- [UNet](https:\u002F\u002Flightning-bolts.readthedocs.io\u002Fen\u002Fstable\u002Fmodels\u002Fconvolutional.html#unet)\n\n###### Reinforcement Learning\n\n- [DQN Loss](https:\u002F\u002Flightning-bolts.readthedocs.io\u002Fen\u002Fstable\u002Flosses.html#dqn-loss)\n- [Double DQN Loss](https:\u002F\u002Flightning-bolts.readthedocs.io\u002Fen\u002Fstable\u002Flosses.html#double-dqn-loss)\n- [Per DQN Loss](https:\u002F\u002Flightning-bolts.readthedocs.io\u002Fen\u002Fstable\u002Flosses.html#per-dqn-loss)\n\n###### GANs\n\n- [Basic GAN](https:\u002F\u002Flightning-bolts.readthedocs.io\u002Fen\u002Fstable\u002Fmodels\u002Fgans.html#basic-gan)\n- [DCGAN](https:\u002F\u002Flightning-bolts.readthedocs.io\u002Fen\u002Fstable\u002Fmodels\u002Fgans.html#dcgan)\n\n###### Classic ML\n\n- [Logistic Regression](https:\u002F\u002Flightning-bolts.readthedocs.io\u002Fen\u002Fstable\u002Fmodels\u002Fclassic_ml.html#logistic-regression)\n- [Linear Regression](https:\u002F\u002Flightning-bolts.readthedocs.io\u002Fen\u002Fstable\u002Fmodels\u002Fclassic_ml.html#linear-regression)\n\n \n \n\n## Continuous Integration\n\nLightning is rigorously tested across multiple CPUs, GPUs and TPUs and against major Python and PyTorch versions.\n\n###### \\*Codecov is > 90%+ but build delays may show less\n\n\u003Cdetails>\n  \u003Csummary>Current build statuses\u003C\u002Fsummary>\n\n\u003Ccenter>\n\n|       System \u002F PyTorch ver.        | 1.13                                                                                                                                                                                                                            | 2.0                                                                                                                                                                                                                             |                                                                                                               2.1                                                                                                               |\n| :--------------------------------: | :-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | :-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: |:-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------:|\n|        Linux py3.9 \\[GPUs\\]        |  |  | [![Build Status](https:\u002F\u002Fdev.azure.com\u002FLightning-AI\u002Flightning\u002F_apis\u002Fbuild\u002Fstatus%2Fpytorch-lightning%20%28GPUs%29?branchName=master)](https:\u002F\u002Fdev.azure.com\u002FLightning-AI\u002Flightning\u002F_build\u002Flatest?definitionId=24&branchName=master) |\n|  Linux (multiple Python versions)  | [![Test PyTorch](https:\u002F\u002Fgithub.com\u002FLightning-AI\u002Flightning\u002Factions\u002Fworkflows\u002Fci-tests-pytorch.yml\u002Fbadge.svg)](https:\u002F\u002Fgithub.com\u002FLightning-AI\u002Flightning\u002Factions\u002Fworkflows\u002Fci-tests-pytorch.yml)                                 | [![Test PyTorch](https:\u002F\u002Fgithub.com\u002FLightning-AI\u002Flightning\u002Factions\u002Fworkflows\u002Fci-tests-pytorch.yml\u002Fbadge.svg)](https:\u002F\u002Fgithub.com\u002FLightning-AI\u002Flightning\u002Factions\u002Fworkflows\u002Fci-tests-pytorch.yml)                                 |                 [![Test PyTorch](https:\u002F\u002Fgithub.com\u002FLightning-AI\u002Flightning\u002Factions\u002Fworkflows\u002Fci-tests-pytorch.yml\u002Fbadge.svg)](https:\u002F\u002Fgithub.com\u002FLightning-AI\u002Flightning\u002Factions\u002Fworkflows\u002Fci-tests-pytorch.yml)                 |\n|   OSX (multiple Python versions)   | [![Test PyTorch](https:\u002F\u002Fgithub.com\u002FLightning-AI\u002Flightning\u002Factions\u002Fworkflows\u002Fci-tests-pytorch.yml\u002Fbadge.svg)](https:\u002F\u002Fgithub.com\u002FLightning-AI\u002Flightning\u002Factions\u002Fworkflows\u002Fci-tests-pytorch.yml)                                 | [![Test PyTorch](https:\u002F\u002Fgithub.com\u002FLightning-AI\u002Flightning\u002Factions\u002Fworkflows\u002Fci-tests-pytorch.yml\u002Fbadge.svg)](https:\u002F\u002Fgithub.com\u002FLightning-AI\u002Flightning\u002Factions\u002Fworkflows\u002Fci-tests-pytorch.yml)                                 |                 [![Test PyTorch](https:\u002F\u002Fgithub.com\u002FLightning-AI\u002Flightning\u002Factions\u002Fworkflows\u002Fci-tests-pytorch.yml\u002Fbadge.svg)](https:\u002F\u002Fgithub.com\u002FLightning-AI\u002Flightning\u002Factions\u002Fworkflows\u002Fci-tests-pytorch.yml)                 |\n| Windows (multiple Python versions) | [![Test PyTorch](https:\u002F\u002Fgithub.com\u002FLightning-AI\u002Flightning\u002Factions\u002Fworkflows\u002Fci-tests-pytorch.yml\u002Fbadge.svg)](https:\u002F\u002Fgithub.com\u002FLightning-AI\u002Flightning\u002Factions\u002Fworkflows\u002Fci-tests-pytorch.yml)                                 | [![Test PyTorch](https:\u002F\u002Fgithub.com\u002FLightning-AI\u002Flightning\u002Factions\u002Fworkflows\u002Fci-tests-pytorch.yml\u002Fbadge.svg)](https:\u002F\u002Fgithub.com\u002FLightning-AI\u002Flightning\u002Factions\u002Fworkflows\u002Fci-tests-pytorch.yml)                                 |                 [![Test PyTorch](https:\u002F\u002Fgithub.com\u002FLightning-AI\u002Flightning\u002Factions\u002Fworkflows\u002Fci-tests-pytorch.yml\u002Fbadge.svg)](https:\u002F\u002Fgithub.com\u002FLightning-AI\u002Flightning\u002Factions\u002Fworkflows\u002Fci-tests-pytorch.yml)                 |\n\n\u003C\u002Fcenter>\n\u003C\u002Fdetails>\n\n \n \n\n## Community\n\nThe lightning community is maintained by\n\n- [10+ core contributors](https:\u002F\u002Flightning.ai\u002Fdocs\u002Fpytorch\u002Flatest\u002Fcommunity\u002Fgovernance.html) who are all a mix of professional engineers, Research Scientists, and Ph.D. students from top AI labs.\n- 800+ community contributors.\n\nWant to help us build Lightning and reduce boilerplate for thousands of researchers? [Learn how to make your first contribution here](https:\u002F\u002Flightning.ai\u002Fdocs\u002Fpytorch\u002Fstable\u002Fgenerated\u002FCONTRIBUTING.html?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme)\n\nLightning is also part of the [PyTorch ecosystem](https:\u002F\u002Fpytorch.org\u002Fecosystem\u002F) which requires projects to have solid testing, documentation and support.\n\n### Asking for help\n\nIf you have any questions please:\n\n1. [Read the docs](https:\u002F\u002Flightning.ai\u002Fdocs?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme).\n1. [Search through existing Discussions](https:\u002F\u002Fgithub.com\u002FLightning-AI\u002Flightning\u002Fdiscussions), or [add a new question](https:\u002F\u002Fgithub.com\u002FLightning-AI\u002Flightning\u002Fdiscussions\u002Fnew)\n1. [Join our discord](https:\u002F\u002Fdiscord.com\u002Finvite\u002FtfXFetEZxv).\n","PyTorch Lightning 是一个用于预训练和微调任意规模AI模型的深度学习框架。它通过简化代码结构，使得用户无需修改现有代码即可在单个或上万个GPU上进行高效训练。其核心功能包括自动处理后向传播、混合精度计算、多GPU及分布式训练等复杂任务，显著减少了开发者的工作量。此外，PyTorch Lightning 还提供了灵活的实验管理和调试工具，增强了模型开发过程中的可维护性和可扩展性。该项目非常适合需要大规模并行训练或希望简化深度学习项目工程负担的研究人员和工程师使用。",2,"2026-06-11 02:48:14","top_all"]