fix: [pre-commit.ci] auto fixes [...]

Author: pre-commit-ci[bot]

torchopt/nn/stateless.py

@@ -84,7 +84,7 @@ def reparametrize(
     module: nn.Module,
     named_tensors: dict[str, torch.Tensor] | Iterable[tuple[str, torch.Tensor]],
     allow_missing: bool = False,
-) -> Generator[nn.Module, None, None]:
+) -> Generator[nn.Module]:
     """Reparameterize the module parameters and/or buffers."""
     if not isinstance(named_tensors, dict):
         named_tensors = dict(named_tensors)

tutorials/2_Visualization.ipynb

@@ -1,5 +1,5 @@
 {
-  "cells": [
+ "cells": [
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -181,8 +181,9 @@
     "# Draw computation graph\n",
     "display(\n",
     "    torchopt.visual.make_dot(\n",
-    "        loss, [net_state_0, net_state_1, {'meta_param': meta_param, 'loss': loss}]\n",
-    "    )\n",
+    "        loss,\n",
+    "        [net_state_0, net_state_1, {'meta_param': meta_param, 'loss': loss}],\n",
+    "    ),\n",
     ")"
    ]
   }

tutorials/3_Meta_Optimizer.ipynb

@@ -1,5 +1,5 @@
 {
-  "cells": [
+ "cells": [
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -200,8 +200,9 @@
     "outer_loss = F.mse_loss(net(x), y)\n",
     "display(\n",
     "    torchopt.visual.make_dot(\n",
-    "        outer_loss, params=[net_state_0, net_state_1, {'x': x, 'outer_loss': outer_loss}]\n",
-    "    )\n",
+    "        outer_loss,\n",
+    "        params=[net_state_0, net_state_1, {'x': x, 'outer_loss': outer_loss}],\n",
+    "    ),\n",
     ")"
    ]
   },
@@ -247,8 +248,9 @@
     "outer_loss = F.mse_loss(net(x), y)\n",
     "display(\n",
     "    torchopt.visual.make_dot(\n",
-    "        outer_loss, params=[net_state_0, net_state_1, {'x': x, 'outer_loss': outer_loss}]\n",
-    "    )\n",
+    "        outer_loss,\n",
+    "        params=[net_state_0, net_state_1, {'x': x, 'outer_loss': outer_loss}],\n",
+    "    ),\n",
     ")"
    ]
   },
@@ -513,21 +515,30 @@
    "source": [
     "functional_adam = torchopt.adam(\n",
     "    lr=torchopt.schedule.linear_schedule(\n",
-    "        init_value=1e-3, end_value=1e-4, transition_steps=10000, transition_begin=2000\n",
-    "    )\n",
+    "        init_value=1e-3,\n",
+    "        end_value=1e-4,\n",
+    "        transition_steps=10000,\n",
+    "        transition_begin=2000,\n",
+    "    ),\n",
     ")\n",
     "\n",
     "adam = torchopt.Adam(\n",
     "    net.parameters(),\n",
     "    lr=torchopt.schedule.linear_schedule(\n",
-    "        init_value=1e-3, end_value=1e-4, transition_steps=10000, transition_begin=2000\n",
+    "        init_value=1e-3,\n",
+    "        end_value=1e-4,\n",
+    "        transition_steps=10000,\n",
+    "        transition_begin=2000,\n",
     "    ),\n",
     ")\n",
     "\n",
     "meta_adam = torchopt.MetaAdam(\n",
     "    net,\n",
     "    lr=torchopt.schedule.linear_schedule(\n",
-    "        init_value=1e-3, end_value=1e-4, transition_steps=10000, transition_begin=2000\n",
+    "        init_value=1e-3,\n",
+    "        end_value=1e-4,\n",
+    "        transition_steps=10000,\n",
+    "        transition_begin=2000,\n",
     "    ),\n",
     ")"
    ]
@@ -610,19 +621,26 @@
     "optim = torchopt.MetaAdam(net, lr=1.0, moment_requires_grad=True, use_accelerated_op=True)\n",
     "\n",
     "net_state_0 = torchopt.extract_state_dict(\n",
-    "    net, by='reference', enable_visual=True, visual_prefix='step0.'\n",
+    "    net,\n",
+    "    by='reference',\n",
+    "    enable_visual=True,\n",
+    "    visual_prefix='step0.',\n",
     ")\n",
     "inner_loss = F.mse_loss(net(x), y)\n",
     "optim.step(inner_loss)\n",
     "net_state_1 = torchopt.extract_state_dict(\n",
-    "    net, by='reference', enable_visual=True, visual_prefix='step1.'\n",
+    "    net,\n",
+    "    by='reference',\n",
+    "    enable_visual=True,\n",
+    "    visual_prefix='step1.',\n",
     ")\n",
     "\n",
     "outer_loss = F.mse_loss(net(x), y)\n",
     "display(\n",
     "    torchopt.visual.make_dot(\n",
-    "        outer_loss, params=[net_state_0, net_state_1, {'x': x, 'outer_loss': outer_loss}]\n",
-    "    )\n",
+    "        outer_loss,\n",
+    "        params=[net_state_0, net_state_1, {'x': x, 'outer_loss': outer_loss}],\n",
+    "    ),\n",
     ")"
    ]
   },

tutorials/4_Stop_Gradient.ipynb

@@ -1,5 +1,5 @@
 {
-  "cells": [
+ "cells": [
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -192,7 +192,7 @@
     "            one_step_net_state,\n",
     "            {'meta_parameter': meta_parameter, 'outer_loss': outer_loss},\n",
     "        ),\n",
-    "    )\n",
+    "    ),\n",
     ")"
    ]
   },
@@ -393,7 +393,7 @@
     "            one_step_net_state,\n",
     "            {'meta_parameter': meta_parameter, 'outer_loss': outer_loss},\n",
     "        ),\n",
-    "    )\n",
+    "    ),\n",
     ")\n",
     "\n",
     "# Outer update\n",
@@ -457,7 +457,9 @@
     "torchopt.stop_gradient(net)\n",
     "torchopt.stop_gradient(optim)\n",
     "one_step_net_state_detached = torchopt.extract_state_dict(\n",
-    "    net, enable_visual=True, visual_prefix='step1.detached.'\n",
+    "    net,\n",
+    "    enable_visual=True,\n",
+    "    visual_prefix='step1.detached.',\n",
     ")\n",
     "\n",
     "# Inner update\n",
@@ -480,7 +482,7 @@
     "            one_step_net_state_detached,\n",
     "            {'meta_parameter': meta_parameter, 'outer_loss': outer_loss},\n",
     "        ),\n",
-    "    )\n",
+    "    ),\n",
     ")"
    ]
   },

tutorials/6_Zero_Order_Differentiation.ipynb

@@ -1,5 +1,5 @@
 {
-  "cells": [
+ "cells": [
   {
    "cell_type": "markdown",
    "id": "8850c832-3b54-4971-8ee0-2cd64b585ea8",
@@ -175,7 +175,11 @@
     "\n",
     "\n",
     "@torchopt.diff.zero_order(\n",
-    "    distribution=distribution, method='forward', argnums=0, num_samples=100, sigma=0.01\n",
+    "    distribution=distribution,\n",
+    "    method='forward',\n",
+    "    argnums=0,\n",
+    "    num_samples=100,\n",
+    "    sigma=0.01,\n",
     ")\n",
     "def forward_process(params, fn, x, y):\n",
     "    y_pred = fn(params, x)\n",