|
Device context management: make sure to set device before calling cuda sync!
|
|
1
|
35
|
March 18, 2026
|
|
How to build PyTorch with a latest LLVM build?
|
|
3
|
957
|
March 18, 2026
|
|
Would helper function recommendations help when prototyping PyTorch projects?
|
|
0
|
14
|
March 13, 2026
|
|
Is There Specific Tensor Operation For That
|
|
0
|
24
|
March 13, 2026
|
|
Easy way to create a FakeTensor with dymbolic shape?
|
|
1
|
32
|
March 11, 2026
|
|
On Counting FLOPs/Energy for the Model
|
|
3
|
88
|
March 11, 2026
|
|
Compatibility Flash attention weights with torch MultiheadAttention
|
|
0
|
33
|
March 11, 2026
|
|
Opdiff: cross-backend PyTorch operator testing + results dashboard
|
|
0
|
23
|
March 8, 2026
|
|
Does torch now offically supported Nvidia Jetson?
|
|
1
|
57
|
March 7, 2026
|
|
Intel GPU support
|
|
0
|
50
|
March 7, 2026
|
|
Title: Interest in ESoC 2026: Contribution and Batch 2 Participation
|
|
1
|
22
|
March 7, 2026
|
|
Compiling earlier PyTorch versions on Blackwell
|
|
3
|
451
|
March 4, 2026
|
|
Cusolver and magma timings for linalg.eigh() (symmetric case)
|
|
1
|
95
|
March 3, 2026
|
|
TorchInductor internals
|
|
0
|
19
|
March 3, 2026
|
|
Error loading "\lib\site-packages\torch\lib\shm.dll" or one of its dependencie
|
|
26
|
33545
|
March 2, 2026
|
|
What is an "unbound" method, such as torch.mul?
|
|
2
|
67
|
February 28, 2026
|
|
Introduction hi everyone how are you all
|
|
0
|
25
|
February 28, 2026
|
|
What do TensorDataset and DataLoader do?
|
|
3
|
35069
|
February 26, 2026
|
|
How to pass float32 into specific model layer but keep the rest of weights float16 (deepspeed)?
|
|
1
|
51
|
February 24, 2026
|
|
Is there an equivalent of jax.lax.scan (eg in torch.func)?
|
|
2
|
2929
|
February 23, 2026
|
|
PR Review request
|
|
0
|
22
|
February 22, 2026
|
|
Pytorch + torchsparse compatibility for rtx5090
|
|
1
|
70
|
February 19, 2026
|
|
Support libtorch for linux arm
|
|
3
|
58
|
February 17, 2026
|
|
Inter-process sharing CUDA Tensor
|
|
1
|
59
|
February 17, 2026
|
|
CUDA support for RTX 50x on Windows
|
|
4
|
1327
|
February 17, 2026
|
|
CTCDecoder returns result.tokens longer than input emissions - why +2 extra tokens?
|
|
5
|
30
|
February 16, 2026
|
|
Fp16_compress_hook casts to FP16 before dividing by world_size — causes NaN with large gradients
|
|
0
|
19
|
February 16, 2026
|
|
Utilising Serialized Torch Models in Apache Spark (scala)
|
|
0
|
28
|
February 14, 2026
|
|
Is there any way to compute multiple jvps without repeatly computing function value?
|
|
5
|
49
|
February 13, 2026
|
|
Is it safe to create new gpu tensor during cuda graph capturing?
|
|
3
|
77
|
February 10, 2026
|