Docs 4/n (#15628)

* remove source-lit

* docs

* docs

* docs

* docs

* ic

* deploy

* deploy

* deploy

* deploy

* deploy

* deploy

* Apply suggestions from code review

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* make build run

Co-authored-by: Jirka Borovec <[email protected]>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Rick Izzo <[email protected]>
(cherry picked from commit d5c0eff)

Author: 4 people

docs/source-app/api_reference/components.rst

@@ -20,6 +20,6 @@ ___________________
 
     ~python.popen.PopenPythonScript
     ~python.tracer.TracerPythonScript
-    ~training.LightningTrainingComponent
+    ~training.LightningTrainerScript
     ~serve.gradio.ServeGradio
     ~serve.serve.ModelInferenceAPI

docs/source-app/api_references.rst

@@ -34,7 +34,7 @@ ___________________
 
     ~python.popen.PopenPythonScript
     ~python.tracer.TracerPythonScript
-    ~training.LightningTrainingComponent
+    ~training.LightningTrainerScript
     ~serve.gradio.ServeGradio
     ~serve.serve.ModelInferenceAPI
 

docs/source-app/index.rst

@@ -6,7 +6,7 @@
 #######################
 Welcome to ⚡ Lightning
 #######################
-Build models and full stack AI apps ⚡ *Lightning fast*.
+Build models, ML components and full stack AI apps ⚡ *Lightning fast*.
 
 .. join_slack::
    :align: left
@@ -22,23 +22,22 @@ Build models and full stack AI apps ⚡ *Lightning fast*.
 
 .. app_card::
    :title: Develop and Train
-   :description: Train an LLM (64 GPUs)
+   :description: Train a model (32 GPUs)
    :width: 280
    :image: https://lightning-ai-docs.s3.amazonaws.com/develop_n_train_v1.jpg
-   :preview: https://lightning.ai
-   :deploy: https://lightning.ai
-   :target: https://apple.com
-   :tags: Model
+   :target: levels/basic/real_lightning_component_implementations.html#ex-pytorch-lightning-trainer
+   :preview: levels/basic/real_lightning_component_implementations.html#ex-pytorch-lightning-trainer
+   :tags: Training
 
 .. app_card::
    :title: Serve and deploy
-   :description: Production-ready stable diffusion server (<2s latency)
+   :description: Production diffusion server (<2s latency)
    :width: 280
    :app_id: HvUwbEG90E
    :image: https://lightning-ai-docs.s3.amazonaws.com/serve_n_deploy_v1.jpg
    :deploy: https://lightning.ai
    :target: https://01gbx4m78rbkpczdf5cpz2hpbh.litng-ai-03.litng.ai/root.api_component/
-   :tags: App
+   :tags: Serving
 
 .. app_card::
    :title: Scale and build a product
@@ -47,7 +46,7 @@ Build models and full stack AI apps ⚡ *Lightning fast*.
    :app_id: HvUwbEG90E
    :image: https://lightning-ai-docs.s3.amazonaws.com/scale_n_build_v1.jpg
    :target: https://lightning.ai/muse
-   :tags: App
+   :tags: AI App
 
 .. raw:: html
 
@@ -56,13 +55,14 @@ Build models and full stack AI apps ⚡ *Lightning fast*.
 
 ----
 
-****************************************
-Build self-contained, modular components
-****************************************
-Lightning is a hyper-minimalistic framework designed to maximize expressivity that
-enables you to build modular, self-contained components and plug them into your existing workflows.
-A Lightning component organizes arbitrary code so it can run on the cloud. A component can train a model, deploy, or even host a web UI.
-The component manages its own infrastructure, cloud costs, networking and more, so you can focus on application logic and not engineering.
+********************************
+Build self-contained, components
+********************************
+Use Lightning, the hyper-minimalistic framework, to build machine learning components that can plug into existing ML workflows.
+A Lightning component organizes arbitrary code to run on the cloud, manage its own infrastructure, cloud costs, networking, and more.
+Focus on component logic and not engineering.
+
+Use components on their own, or compose them into full-stack AI apps with our next-generation Lightning orchestrator.
 
 .. raw:: html
 
@@ -78,6 +78,10 @@ The component manages its own infrastructure, cloud costs, networking and more,
 
 .. include:: ./levels/basic/hero_components.rst
 
+|
+
+Components run the same on the cloud and locally on your choice of hardware.
+
 .. lit_tabs::
    :code_files: landing_app_run.bash
    :highlights: 5

docs/source-app/landing_app_run.bash

@@ -1,5 +1,5 @@
 # install lightning
 pip install lightning
 
-# run the app
-lightning run app app.py --cloud
+# run the app on the --cloud (--setup installs deps automatically)
+lightning run app app.py --setup --cloud

docs/source-app/levels/basic/build_a_lightning_component.rst

@@ -14,13 +14,12 @@ Level 1: Package code in a lightning component
 *********************************
 Why you need Lightning components
 *********************************
-A Lightning component organizes a piece of code into a self-contained, modular component that
-can be integrated into your existing workflows or assembled to form a Lightning app.
-A Lightning component manages its own infrastructure, auto-scaling, cost management, and more, so you
-can focus on the program logic and not the cloud engineering.
+A Lightning component is a self-contained, modular machine-learning component
+that you can plug into your existing ML workflows. A Lightning component organizes arbitrary code so it can run on the cloud, manages
+its own infrastructure, cloud costs, networking and more. Connect components using your current workflow management tools or
+our `next-generation reactive orchestrator <../intermediate/index.html>`_.
 
-Components run on the cloud or your laptop without code changes 🤯🤯. Connect components using your current workflow management tools or use
-Lightning apps to build powerful sequential AND reactive workflows.
+Components run on the cloud or your laptop without code changes 🤯🤯.
 
 .. raw:: html
 
@@ -109,7 +108,7 @@ First, install Lightning.
 **************************
 Build your first component
 **************************
-A Lightning component organizes Python code into a self-contained module so it can run on the cloud.
+A Lightning component organizes arbitrary code so it can run on the cloud, manages its own infrastructure, cloud costs, networking and more
 
 **Run one of these components!**
 
@@ -119,12 +118,7 @@ A Lightning component organizes Python code into a self-contained module so it c
 
 Components run the same on the cloud and locally on your choice of hardware.
 
-.. lit_tabs::
-   :titles: Lightning Cloud (fully-managed); Your AWS account; Your own hardware
-   :code_files: ./hello_components/code_run_cloud.bash; ./hello_components/code_run_cloud_yours.bash; ./hello_components/code_run_local.bash
-   :tab_rows: 4
-   :highlights: ; 5; 0
-   :height: 195px
+.. include:: /levels/basic/hero_run_setup.rst
 
 ----
 

docs/source-app/levels/basic/hello_components/code_run_cloud_setup.bash

@@ -0,0 +1 @@
+lightning run app app.py --setup --cloud

docs/source-app/levels/basic/hello_components/code_run_cloud_yours_setup.bash

@@ -0,0 +1,5 @@
+# first create a cluster (creation could take ~30 minutes)
+lightning create cluster pikachu --provider aws --role-arn arn:aws:iam::1234567890:role/lai-byoc
+
+# run on that cluster
+lightning run app app.py --setup --cloud pikachu

docs/source-app/levels/basic/hello_components/code_run_local_setup.bash

@@ -0,0 +1 @@
+lightning run app app.py --setup

docs/source-app/levels/basic/hello_components/deploy_model.py

@@ -1,15 +1,31 @@
-# A hello world component
-# app.py
+# !pip install torchvision
 import lightning as L
+from lightning.app.components.serve import PythonServer, Image, Number
+import base64, io, torchvision, torch
+from PIL import Image as PILImage
 
 
-class YourComponent(L.LightningWork):
-   def run(self):
-      print('RUN ANY PYTHON CODE HERE')
+class PyTorchServer(PythonServer):
+    def setup(self):
+        self._model = torchvision.models.resnet18(pretrained=True)
+        self._device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
+        self._model.to(self._device)
 
+    def predict(self, request):
+        image = base64.b64decode(request.image.encode("utf-8"))
+        image = PILImage.open(io.BytesIO(image))
+        transforms = torchvision.transforms.Compose([
+            torchvision.transforms.Resize(224),
+            torchvision.transforms.ToTensor(),
+            torchvision.transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
+        ])
+        image = transforms(image)
+        image = image.to(self._device)
+        prediction = self._model(image.unsqueeze(0))
+        return {"prediction": prediction.argmax().item()}
 
 
-# run on a cloud machine
-compute = L.CloudCompute("cpu")
-worker = YourComponent(cloud_compute=compute)
-app = L.LightningApp(worker)
+component = PyTorchServer(
+   input_type=Image, output_type=Number, cloud_compute=L.CloudCompute('gpu')
+)
+app = L.LightningApp(component)

docs/source-app/levels/basic/hello_components/pl_multinode.py

@@ -0,0 +1,20 @@
+# app.py
+import lightning as L
+from lightning.app.components import PyTorchLightningMultiNode
+from lightning.pytorch.demos.boring_classes import BoringModel
+
+
+class LightningTrainerDistributed(L.LightningWork):
+    @staticmethod
+    def run():
+        model = BoringModel()
+        trainer = L.Trainer(max_epochs=10, strategy="ddp")
+        trainer.fit(model)
+
+# 8 GPU: (2 nodes of 4 x v100)
+component = PyTorchLightningMultiNode(
+    LightningTrainerDistributed,
+    num_nodes=2,
+    cloud_compute=L.CloudCompute("gpu-fast-multi"), # 4 x v100
+)
+app = L.LightningApp(component)