Merge branch 'Version2.0' into main

Signed-off-by: Tom Freudenberg <[email protected]>
boschresearch · Jun 20, 2024 · 79a4b15 · 79a4b15
2 parents 3b99722 + c055588
commit 79a4b15
Show file tree

Hide file tree

Showing 45 changed files with 23,732 additions and 1,001 deletions.
diff --git a/.gitignore b/.gitignore
@@ -16,9 +16,9 @@ __pycache__/*
 
 # Log folders
 **/lightning_logs/**
-**/fluid_logs/**
-**/Multiscale/**
 **/bosch/**
+
+**/fluid_logs/**
 # Project files
 .ropeproject
 .project

diff --git a/README.rst b/README.rst
@@ -80,32 +80,24 @@ Installation
 ============
 TorchPhysics reqiueres the follwing dependencies to be installed: 
 
-- PyTorch_ >= 1.7.1, < 2.0.0
-- `PyTorch Lightning`_ >= 1.3.4, < 2.0.0
+- PyTorch_ >= 2.0.0
+- `PyTorch Lightning`_ >= 2.0.0
 - Numpy_ >= 1.20.2
 - Matplotlib_ >= 3.0.0
 - Scipy_ >= 1.6.3
 
-Installing TorchPhysics with ``pip``, automatically downloads everything that is needed:
+To install the TorchPhysics version that is compatible with PyTorch>=2.0.0
+you have to copy this branch and install it manually, either by
 
 .. code-block:: python
 
-  pip install torchphysics
+  pip install "git+https://github.com/boschresearch/torchphysics@Version2.0"
 
-Additionally, to use the ``Shapely`` and ``Trimesh`` functionalities, install the library 
-with the option ``all``:
+or by
 
 .. code-block:: python
 
-  pip install torchphysics[all]
-
-
-If you want to add functionalities or modify the code. We recommend copying the 
-repository and installing it locally:
-
-.. code-block:: python
-
-  git clone https://github.com/boschresearch/torchphysics 
+  git clone https://github.com/boschresearch/torchphysics@Version2.0 
   cd path_to_torchphysics_folder
   pip install .[all]
 

diff --git a/docs/requirements.txt b/docs/requirements.txt
@@ -5,8 +5,8 @@ protobuf~=3.19.0 # fix github test and docu creation
 sphinx>=3.2.1
 sphinx_rtd_theme
 
-torch>=1.7.1
-pytorch-lightning>=1.3.4
+torch>=2.0.0
+pytorch-lightning>=2.0.0
 numpy>=1.20.2
 matplotlib>=3.4.2
 trimesh>=3.9.19

diff --git a/docs/tutorial/solve_pde.rst b/docs/tutorial/solve_pde.rst
@@ -169,11 +169,12 @@ Afterwards we switch to LBFGS:
 
     solver = tp.solver.Solver(train_conditions=[bound_cond, pde_cond], optimizer_setting=optim)
 
-    trainer = pl.Trainer(gpus=1,
-                         max_steps=3000, # number of training steps
-                         logger=False,
-                         benchmark=True,
-                         checkpoint_callback=False)
+    trainer = pl.Trainer(devices=1, accelerator="gpu",
+                        num_sanity_val_steps=0,
+                        benchmark=True,
+                        max_steps=3000, 
+                        logger=False, 
+                        enable_checkpointing=False)
                         
     trainer.fit(solver)
 

diff --git a/docs/tutorial/solver_info.rst b/docs/tutorial/solver_info.rst
@@ -22,15 +22,20 @@ the basic trainer is defined as follows:
   
   import pytorch_lightning as pl
 
-  trainer = pl.Trainer(gpus=1, max_steps=4000, check_val_every_n_epoch=10)
+  trainer = pl.Trainer(devices=1, accelerator="gpu", 
+                        max_steps=4000, check_val_every_n_epoch=10)
   trainer.fit(solver) # start training
 
 Some important keywords are:
 
-- **gpus**: The number of GPUs that should be used. If only one or more CPUs are available
-  set ```gpus=None``. Depending on the operating system, the GPUs may have to be further
-  specified beforehand via ``os.environ`` or other ways. There are more different possiblities to 
-  specify the used device, see the above mentionde documentation.
+- **devices**: The number of devices that should be used. 
+- **accelerator**: On what kind of device the network will be trained.
+  Usually, one needs at least one GPU to train a neural network sufficiently well. If only 
+  CPUs are available set ``"cpu"``. 
+  Depending on the operating system, the GPUs may have to be further
+  specified beforehand via ``os.environ`` or other ways. 
+  There are more different possibilities to specify the used device, 
+  see the above-mentioned documentation.
 - **max_steps**: The maximum number of training iterations. In each iteration
   all defined training conditions will be evaluated 
   (e.g. points sampled, model output computed, residuals evaluated, etc.) and a 

diff --git a/examples/deeponet/inverse_ode.ipynb b/examples/deeponet/inverse_ode.ipynb
@@ -258,13 +258,12 @@
     "\n",
     "solver = tp.solver.Solver([data_condition], optimizer_setting=optim)\n",
     "\n",
-    "trainer = pl.Trainer(gpus='-1' if torch.cuda.is_available() else None,\n",
+    "trainer = pl.Trainer(devices=1, accelerator=\"gpu\",\n",
     "                     num_sanity_val_steps=0,\n",
     "                     benchmark=True,\n",
-    "                     max_steps=10000,\n",
-    "                     logger=False,\n",
-    "                     checkpoint_callback=False\n",
-    "                     )\n",
+    "                     max_steps=10000, \n",
+    "                     logger=False, \n",
+    "                     enable_checkpointing=False)\n",
     "\n",
     "trainer.fit(solver)"
    ]
@@ -327,13 +326,12 @@
     "\n",
     "solver = tp.solver.Solver([data_condition], optimizer_setting=optim)\n",
     "\n",
-    "trainer = pl.Trainer(gpus='-1' if torch.cuda.is_available() else None,\n",
+    "trainer = pl.Trainer(devices=1, accelerator=\"gpu\",\n",
     "                     num_sanity_val_steps=0,\n",
     "                     benchmark=True,\n",
-    "                     max_steps=1000,\n",
-    "                     logger=False,\n",
-    "                     checkpoint_callback=False\n",
-    "                     )\n",
+    "                     max_steps=1000, \n",
+    "                     logger=False, \n",
+    "                     enable_checkpointing=False)\n",
     "\n",
     "trainer.fit(solver)"
    ]

diff --git a/examples/deeponet/ode.ipynb b/examples/deeponet/ode.ipynb
@@ -191,13 +191,12 @@
     "\n",
     "solver = tp.solver.Solver([ode_cond, initial_cond], optimizer_setting=optim)\n",
     "\n",
-    "trainer = pl.Trainer(gpus='-1' if torch.cuda.is_available() else None,\n",
+    "trainer = pl.Trainer(devices=1, accelerator=\"gpu\",\n",
     "                     num_sanity_val_steps=0,\n",
     "                     benchmark=True,\n",
-    "                     max_steps=5000,\n",
-    "                     logger=False,\n",
-    "                     checkpoint_callback=False\n",
-    "                     )\n",
+    "                     max_steps=5000, \n",
+    "                     logger=False, \n",
+    "                     enable_checkpointing=False)\n",
     "\n",
     "trainer.fit(solver)"
    ]
@@ -284,11 +283,12 @@
     "\n",
     "solver = tp.solver.Solver(train_conditions=[ode_cond, initial_cond], optimizer_setting=optim)\n",
     "\n",
-    "trainer = pl.Trainer(gpus=1,\n",
-    "                     max_steps=3000, \n",
-    "                     logger=False,\n",
+    "trainer = pl.Trainer(devices=1, accelerator=\"gpu\",\n",
+    "                     num_sanity_val_steps=0,\n",
     "                     benchmark=True,\n",
-    "                     checkpoint_callback=False)\n",
+    "                     max_steps=3000, \n",
+    "                     logger=False, \n",
+    "                     enable_checkpointing=False)\n",
     "                     \n",
     "trainer.fit(solver)"
    ]

diff --git a/examples/deeponet/oscillator.ipynb b/examples/deeponet/oscillator.ipynb
@@ -8,8 +8,8 @@
    "source": [
     "import os\n",
     "os.environ[\"CUDA_VISIBLE_DEVICES\"] = \"3\"\n",
-    "import torch\n",
     "import torchphysics as tp\n",
+    "import torch\n",
     "import pytorch_lightning as pl"
    ]
   },
@@ -219,13 +219,12 @@
    "source": [
     "solver = tp.solver.Solver([ode_cond])\n",
     "\n",
-    "trainer = pl.Trainer(gpus='-1' if torch.cuda.is_available() else None,\n",
+    "trainer = pl.Trainer(devices=1, accelerator=\"gpu\",\n",
     "                     num_sanity_val_steps=0,\n",
     "                     benchmark=True,\n",
-    "                     max_steps=2000,\n",
-    "                     logger=False,\n",
-    "                     checkpoint_callback=False\n",
-    "                     )\n",
+    "                     max_steps=30000, \n",
+    "                     logger=False, \n",
+    "                     enable_checkpointing=False)\n",
     "\n",
     "trainer.fit(solver)"
    ]
@@ -311,12 +310,12 @@
     "\n",
     "solver = tp.solver.Solver(train_conditions=[ode_cond], optimizer_setting=optim)\n",
     "\n",
-    "trainer = pl.Trainer(gpus=1,\n",
-    "                     max_steps=2500, \n",
-    "                     logger=False,\n",
+    "trainer = pl.Trainer(devices=1, accelerator=\"gpu\",\n",
+    "                     num_sanity_val_steps=0,\n",
     "                     benchmark=True,\n",
-    "                     checkpoint_callback=False)\n",
-    "                     \n",
+    "                     max_steps=2500, \n",
+    "                     logger=False, \n",
+    "                     enable_checkpointing=False)\n",
     "trainer.fit(solver)"
    ]
   },
@@ -454,10 +453,10 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.9.15"
+   "version": "3.11.7"
   },
   "orig_nbformat": 4
  },
  "nbformat": 4,
  "nbformat_minor": 2
-}
+}
diff --git a/examples/deepritz/corner_pde.ipynb b/examples/deepritz/corner_pde.ipynb
@@ -161,12 +161,12 @@
    "source": [
     "import pytorch_lightning as pl\n",
     "\n",
-    "trainer = pl.Trainer(gpus=1, # or None if CPU is used\n",
-    "                     max_steps=5000, # number of training steps\n",
-    "                     logger=False,\n",
+    "trainer = pl.Trainer(devices=1, accelerator=\"gpu\",\n",
+    "                     num_sanity_val_steps=0,\n",
     "                     benchmark=True,\n",
-    "                     checkpoint_callback=False)\n",
-    "                     \n",
+    "                     max_steps=5000, \n",
+    "                     logger=False, \n",
+    "                     enable_checkpointing=False)\n",
     "trainer.fit(solver)"
    ]
   },
@@ -245,11 +245,12 @@
     "bound_cond.sampler = bound_cond.sampler.make_static() \n",
     "solver = tp.solver.Solver(train_conditions=[bound_cond, pde_cond], optimizer_setting=optim)\n",
     "\n",
-    "trainer = pl.Trainer(gpus=1,\n",
-    "                     max_steps=3000, # number of training steps\n",
-    "                     logger=False,\n",
+    "trainer = pl.Trainer(devices=1, accelerator=\"gpu\",\n",
+    "                     num_sanity_val_steps=0,\n",
     "                     benchmark=True,\n",
-    "                     checkpoint_callback=False)\n",
+    "                     max_steps=3000, \n",
+    "                     logger=False, \n",
+    "                     enable_checkpointing=False)\n",
     "                     \n",
     "trainer.fit(solver)"
    ]

diff --git a/examples/deepritz/poisson-equation.ipynb b/examples/deepritz/poisson-equation.ipynb
@@ -14,8 +14,11 @@
    "metadata": {},
    "outputs": [],
    "source": [
+    "import os\n",
+    "os.environ[\"CUDA_VISIBLE_DEVICES\"] = \"0\"\n",
     "import torch\n",
-    "import torchphysics as tp"
+    "import torchphysics as tp\n",
+    "import pytorch_lightning as pl"
    ]
   },
   {
@@ -238,15 +241,12 @@
     "solver = tp.solver.Solver([pde_condition,\n",
     "                           boundary_condition])\n",
     "\n",
-    "import pytorch_lightning as pl\n",
-    "import os\n",
-    "os.environ[\"CUDA_VISIBLE_DEVICES\"] = \"0\"\n",
-    "\n",
-    "trainer = pl.Trainer(gpus=None, # or None for CPU\n",
-    "                     max_steps=2000,\n",
-    "                     logger=False,\n",
+    "trainer = pl.Trainer(devices=1, accelerator=\"gpu\",\n",
+    "                     num_sanity_val_steps=0,\n",
     "                     benchmark=True,\n",
-    "                     checkpoint_callback=False)\n",
+    "                     max_steps=2000, \n",
+    "                     logger=False, \n",
+    "                     enable_checkpointing=False)\n",
     "trainer.fit(solver)"
    ]
   },

diff --git a/examples/oscillation/duffing_nonlinear.ipynb b/examples/oscillation/duffing_nonlinear.ipynb