Merge branch 'main' into move-back-to-stable-binaries-2.2

Author: svekars

advanced_source/coding_ddpg.py

@@ -63,16 +63,25 @@
 #      %%bash
 #      pip3 install torchrl mujoco glfw
 
-import torchrl
-import torch
-import tqdm
-from typing import Tuple
-
 # sphinx_gallery_start_ignore
 import warnings
 warnings.filterwarnings("ignore")
+import multiprocessing
+# TorchRL prefers spawn method, that restricts creation of  ``~torchrl.envs.ParallelEnv`` inside
+# `__main__` method call, but for the easy of reading the code switch to fork
+# which is also a default spawn method in Google's Colaboratory
+try:
+    multiprocessing.set_start_method("fork")
+except RuntimeError:
+    assert multiprocessing.get_start_method() == "fork"
 # sphinx_gallery_end_ignore
 
+
+import torchrl
+import torch
+import tqdm
+from typing import Tuple
+
 ###############################################################################
 # We will execute the policy on CUDA if available
 device =  torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
@@ -1219,6 +1228,6 @@ def ceil_div(x, y):
 # 
 # To iterate further on this loss module we might consider:
 # 
-# - Using `@dispatch` (see `[Feature] Distpatch IQL loss module <https://github.com/pytorch/rl/pull/1230>`_.
+# - Using `@dispatch` (see `[Feature] Distpatch IQL loss module <https://github.com/pytorch/rl/pull/1230>`_.)
 # - Allowing flexible TensorDict keys.
 # 

index.rst

@@ -3,15 +3,11 @@ Welcome to PyTorch Tutorials
 
 What's new in PyTorch tutorials?
 
-* `Getting Started with Distributed Checkpoint (DCP) <https://pytorch.org/tutorials/recipes/distributed_checkpoint_recipe.html>`__
-* `torch.export Tutorial <https://pytorch.org/tutorials/intermediate/torch_export_tutorial.html>`__
-* `Facilitating New Backend Integration by PrivateUse1 <https://pytorch.org/tutorials/advanced/privateuseone.html>`__
-* `(prototype) Accelerating BERT with semi-structured (2:4) sparsity <https://pytorch.org/tutorials/prototype/semi_structured_sparse.html>`__
-* `(prototype) PyTorch 2 Export Quantization-Aware Training (QAT) <https://pytorch.org/tutorials/prototype/pt2e_quant_qat.html>`__
-* `(prototype) PyTorch 2 Export Post Training Quantization with X86 Backend through Inductor <https://pytorch.org/tutorials/prototype/pt2e_quant_ptq_x86_inductor.html>`__
-* `(prototype) Inductor C++ Wrapper Tutorial <https://pytorch.org/tutorials/prototype/inductor_cpp_wrapper_tutorial.html>`__
-* `How to save memory by fusing the optimizer step into the backward pass <https://pytorch.org/tutorials/intermediate/optimizer_step_in_backward_tutorial.html>`__
-* `Tips for Loading an nn.Module from a Checkpoint <https://pytorch.org/tutorials/recipes/recipes/module_load_state_dict_tips.html>`__
+* `PyTorch Inference Performance Tuning on AWS Graviton Processors <https://pytorch.org/tutorials/recipes/inference_tuning_on_aws_graviton.html>`__
+* `Using TORCH_LOGS python API with torch.compile <https://pytorch.org/tutorials/recipes/torch_logs.html>`__
+* `PyTorch 2 Export Quantization with X86 Backend through Inductor <https://pytorch.org/tutorials/prototype/pt2e_quant_x86_inductor.html>`__
+* `Getting Started with DeviceMesh <https://pytorch.org/tutorials/recipes/distributed_device_mesh.html>`__
+* `Compiling the optimizer with torch.compile <https://pytorch.org/tutorials/recipes/compiling_optimizer.html>`__
 
 
 .. raw:: html

recipes_source/compiling_optimizer.rst

@@ -86,3 +86,9 @@ Sample Results:
 
 * Eager runtime: 747.2437149845064us
 * Compiled runtime: 392.07384741178us
+
+See Also
+~~~~~~~~~
+
+* For an in-depth technical overview, see
+`Compiling the optimizer with PT2 <https://dev-discuss.pytorch.org/t/compiling-the-optimizer-with-pt2/1669>`__