|
Pytorch to Triton for Non-GPU Devices
|
|
4
|
317
|
April 24, 2024
|
|
Understanding CUDAGraph Trees
|
|
4
|
620
|
April 11, 2024
|
|
State of PT2 OSS Issues: Q1 2024
|
|
0
|
199
|
April 10, 2024
|
|
Inductor CUDA Backend
|
|
1
|
440
|
April 4, 2024
|
|
TorchDynamo Update 9: Making DDP Work with TorchDynamo
|
|
8
|
11580
|
November 27, 2023
|
|
Difference torch dynamo and torch script
|
|
0
|
245
|
March 27, 2024
|
|
A TorchDynamo trace time ablation study
|
|
0
|
442
|
March 22, 2024
|
|
FMAs (and softmax (and floating point)) considered harmful
|
|
0
|
357
|
March 20, 2024
|
|
Supporting mutations in torch.export.export
|
|
2
|
729
|
March 20, 2024
|
|
Meaning of strict=False in torch.export.export
|
|
0
|
148
|
March 20, 2024
|
|
Export sub-graphs at the Aten IR level
|
|
2
|
246
|
March 14, 2024
|
|
PyTorch 2.0 User Empathy Day Recap
|
|
2
|
403
|
March 13, 2024
|
|
Torch.compile() + FSDP - Dec 8th
|
|
2
|
1965
|
March 8, 2024
|
|
TorchInductor Update 7: key optimizations with CPU backend in PyTorch 2.2 release
|
|
4
|
580
|
March 8, 2024
|
|
Custom cuda extension support in Inductor
|
|
8
|
395
|
March 7, 2024
|
|
Trying to understand flow for compilation
|
|
1
|
246
|
March 7, 2024
|
|
Is there a plan for FX in a C++ IR?
|
|
2
|
465
|
February 28, 2024
|
|
PyTorch/XLA 2.2 Release Dev Update
|
|
0
|
684
|
February 23, 2024
|
|
How is pattern matching in inductor/fx implemented?
|
|
7
|
1252
|
February 14, 2024
|
|
Example inputs to compilers are now fake tensors
|
|
9
|
2425
|
February 13, 2024
|
|
TorchDynamo Update 4: LazyTensor & nvFuser Experiments
|
|
4
|
4282
|
February 9, 2024
|
|
Connecting PyTorch sparse tensors with MLIR
|
|
3
|
673
|
February 9, 2024
|
|
Cannot mutate inputs in aot_export workflow
|
|
0
|
173
|
February 7, 2024
|
|
Developer docs for PyTorch inductor?
|
|
3
|
738
|
February 6, 2024
|
|
Compiling the optimizer with PT2
|
|
8
|
2131
|
January 29, 2024
|
|
FYI: Many Dynamo tests were erroneously passing
|
|
0
|
312
|
January 29, 2024
|
|
`compile_autograd`
|
|
0
|
220
|
January 26, 2024
|
|
Micro-optimizations for the most micro of benchmarks
|
|
0
|
572
|
January 25, 2024
|
|
`torch.compile` `AOTAutograd` backwards _inductor function
|
|
0
|
288
|
January 23, 2024
|
|
[GUIDE] Getting C++ custom ops to work with torch.compile
|
|
3
|
1252
|
January 16, 2024
|