From 4fa2e2e506db01aedbb4df7a619f22a3b46c24a0 Mon Sep 17 00:00:00 2001
From: thomas chaton <thomas@grid.ai>
Date: Tue, 8 Nov 2022 14:37:15 +0000
Subject: [PATCH 1/2] remove lite

---
 examples/app_multi_node/README.md    | 10 ----------
 examples/app_multi_node/train_any.py | 22 ----------------------
 2 files changed, 32 deletions(-)
 delete mode 100644 examples/app_multi_node/train_any.py

diff --git a/examples/app_multi_node/README.md b/examples/app_multi_node/README.md
index 23e7afa23d68e..02871dc77d06b 100644
--- a/examples/app_multi_node/README.md
+++ b/examples/app_multi_node/README.md
@@ -18,16 +18,6 @@ or you can use the built-in component for it.
 lightning run app train_pytorch_spawn.py
 ```
 
-## Multi Node with raw PyTorch + Lite
-
-You can run the multi-node raw PyTorch and Lite by running the following commands.
-
-```bash
-lightning run app train_lite.py
-```
-
-Using Lite, you retain control over your loops while accessing in a minimal way all Lightning distributed strategies.
-
 ## Multi Node with PyTorch Lightning
 
 Lightning supports running PyTorch Lightning from a script or within a Lightning Work.
diff --git a/examples/app_multi_node/train_any.py b/examples/app_multi_node/train_any.py
deleted file mode 100644
index 5dfb947134dcd..0000000000000
--- a/examples/app_multi_node/train_any.py
+++ /dev/null
@@ -1,22 +0,0 @@
-import lightning as L
-from lightning.app.components import MultiNode
-
-
-class AnyDistributedComponent(L.LightningWork):
-    def run(
-        self,
-        main_address: str,
-        main_port: int,
-        num_nodes: int,
-        node_rank: int,
-    ):
-        print(f"ADD YOUR DISTRIBUTED CODE: {main_address} {main_port} {num_nodes} {node_rank}.")
-
-
-app = L.LightningApp(
-    MultiNode(
-        AnyDistributedComponent,
-        num_nodes=2,
-        cloud_compute=L.CloudCompute("gpu"),
-    )
-)

From 0d2db5d28e1d645e7171d0b921316baf5eec5c24 Mon Sep 17 00:00:00 2001
From: thomas chaton <thomas@grid.ai>
Date: Tue, 8 Nov 2022 14:40:57 +0000
Subject: [PATCH 2/2] update

---
 examples/app_multi_node/train_any.py  | 22 ++++++++++++++++
 examples/app_multi_node/train_lite.py | 38 ---------------------------
 2 files changed, 22 insertions(+), 38 deletions(-)
 create mode 100644 examples/app_multi_node/train_any.py
 delete mode 100644 examples/app_multi_node/train_lite.py

diff --git a/examples/app_multi_node/train_any.py b/examples/app_multi_node/train_any.py
new file mode 100644
index 0000000000000..5dfb947134dcd
--- /dev/null
+++ b/examples/app_multi_node/train_any.py
@@ -0,0 +1,22 @@
+import lightning as L
+from lightning.app.components import MultiNode
+
+
+class AnyDistributedComponent(L.LightningWork):
+    def run(
+        self,
+        main_address: str,
+        main_port: int,
+        num_nodes: int,
+        node_rank: int,
+    ):
+        print(f"ADD YOUR DISTRIBUTED CODE: {main_address} {main_port} {num_nodes} {node_rank}.")
+
+
+app = L.LightningApp(
+    MultiNode(
+        AnyDistributedComponent,
+        num_nodes=2,
+        cloud_compute=L.CloudCompute("gpu"),
+    )
+)
diff --git a/examples/app_multi_node/train_lite.py b/examples/app_multi_node/train_lite.py
deleted file mode 100644
index ed9777a1064f6..0000000000000
--- a/examples/app_multi_node/train_lite.py
+++ /dev/null
@@ -1,38 +0,0 @@
-import torch
-
-import lightning as L
-from lightning.app.components import LiteMultiNode
-from lightning.lite import LightningLite
-
-
-class LitePyTorchDistributed(L.LightningWork):
-    @staticmethod
-    def run():
-        # 1. Create LightningLite.
-        lite = LightningLite(strategy="ddp", precision="bf16")
-
-        # 2. Prepare distributed model and optimizer.
-        model = torch.nn.Linear(32, 2)
-        optimizer = torch.optim.SGD(model.parameters(), lr=0.01)
-        model, optimizer = lite.setup(model, optimizer)
-        criterion = torch.nn.MSELoss()
-
-        # 3. Train the model for 50 steps.
-        for step in range(50):
-            model.zero_grad()
-            x = torch.randn(64, 32).to(lite.device)
-            output = model(x)
-            loss = criterion(output, torch.ones_like(output))
-            print(f"global_rank: {lite.global_rank} step: {step} loss: {loss}")
-            lite.backward(loss)
-            optimizer.step()
-
-
-# Run over 2 nodes of 4 x V100
-app = L.LightningApp(
-    LiteMultiNode(
-        LitePyTorchDistributed,
-        cloud_compute=L.CloudCompute("gpu-fast-multi"),  # 4 x V100
-        num_nodes=2,
-    )
-)