torch.compile

torch._dynamo PyTorch’s compiler frontend which can output subgraphs when programs aren’t traceable. FX (Functional Transformations) FX is a toolkit for developers to use to transform nn.Module instances. FX consists of three main components: a symbolic tracer, an intermediate representation , and Python code generation . torch._inductor The default compiler in torch.compile, builds loops schedules and codegens triton or C++ kernels.

Topic	Replies	Views	Activity
About the torch.compile category torch.compile	0	1244	January 9, 2023
Choice of torch.compile vs. triton torch.compile	7	7647	August 29, 2025
Reducing compiled call overhead torch.compile	0	5	August 29, 2025
Dynamo logs generated by torch.distributed torch.compile	2	30	August 27, 2025
PyTorch Error: Compiler: cl is not found torch._inductor	3	884	August 27, 2025
How can we reliably figured out what the parameters of a generated PTX kernel are with torch.compile()? torch.compile	0	17	August 25, 2025
Module 'torch.library' has no attribute 'custom_op' torch.compile	1	206	August 23, 2025
C++/Cuda / aot.compile and cuda graph torch.compile	2	537	August 23, 2025
Training with flex attention is extremely slow due to torch.compile settings torch.compile	0	21	August 22, 2025
Is it possible that node's name can change during torch.compile? torch.compile	0	10	August 22, 2025
Torch.compile Numpy code throws mean() arguments error torch.compile	0	8	August 21, 2025
Using RNG Generator in torch.compile torch.compile	2	28	August 20, 2025
How to turn off inlining / force materialization in TorchInductor during torch.compile? torch._inductor	0	13	August 16, 2025
Some questions about torch.compile torch.compile	0	18	August 15, 2025
`torch.compile` (w/ Torch Inductor) benchmarks/models for Multi GPU torch._inductor	0	26	August 7, 2025
CUDA Graph Error with Residual Connections in `torch.compile` (RuntimeError: accessing tensor output of CUDAGraphs) torch.compile	0	33	August 1, 2025
The CUDA kernel produces different results when running in CUDA Graph mode compared to non-CUDA Graph mode torch.compile	0	16	July 30, 2025
Skip dynamo when using 'cudagraph' backend torch.compile	0	31	July 28, 2025
Getting the fx graph of submodules, instead of 'call_module' nodes? torch.compile	1	731	July 24, 2025
Using megacache vs saving /tmp/torchinductor_root torch.compile	0	59	July 21, 2025
Is there any way to save multiple graphs from same model without redundantly saving their state dicts? torch._dynamo	2	50	July 19, 2025
_inductor/schduler.py line#3161 torch.compile	0	19	July 6, 2025
Why pytorch use triton as backend of inductor? torch.compile	2	110	July 3, 2025
Disabling guards generation using dynamo based export torch._dynamo	1	46	June 30, 2025
Memory Footprint of Eager Mode vs Torch Compiled (Default, with CUDA Graph, and CUDA Graph Dynamic) torch.compile	4	1709	June 27, 2025
Torch.compile _softmax in MultiheadAttention does not return same value as eager torch.compile	4	53	June 26, 2025
Graph exported by torch.onnx.dynamo_export torch.compile	2	503	June 24, 2025
Issues with torch.compile in DistributedModelParallel (DMP) torch.compile	1	29	June 19, 2025
Issues with PyTorch while_loop operator when exporting the torch model to ExecuTorch torch._dynamo	1	50	June 19, 2025
Building from GIT error: PyTorch needs CUDNN-8.5 or above but found 8.2.4 torch.compile	4	72	June 17, 2025