|
Are MPS operators rewritten?
|
|
2
|
44
|
February 18, 2026
|
|
Extracting int8 weights and other quant params after convert_pt2e
|
|
1
|
39
|
February 17, 2026
|
|
XNNPACKQuantizer.set_module_name() not working as expected
|
|
2
|
40
|
February 17, 2026
|
|
Limitations of Int8 QAT for Linear Layers
|
|
1
|
38
|
February 17, 2026
|
|
Severe training slowdown after unfold-based patchification
|
|
5
|
66
|
February 17, 2026
|
|
Question About Backward–ReduceScatter Overlap in FSDP Figure 5
|
|
2
|
29
|
February 17, 2026
|
|
Support libtorch for linux arm
|
|
3
|
42
|
February 17, 2026
|
|
Inter-process sharing CUDA Tensor
|
|
1
|
30
|
February 17, 2026
|
|
CUDA support for RTX 50x on Windows
|
|
4
|
1187
|
February 17, 2026
|
|
RPN+ROI architectiure (Faster R-CNN)
|
|
1
|
838
|
February 16, 2026
|
|
Nested torch.compile-d function calls with different options / CUDA graph options
|
|
0
|
24
|
February 16, 2026
|
|
CTCDecoder returns result.tokens longer than input emissions - why +2 extra tokens?
|
|
5
|
29
|
February 16, 2026
|
|
Fp16_compress_hook casts to FP16 before dividing by world_size — causes NaN with large gradients
|
|
0
|
15
|
February 16, 2026
|
|
Utilising Serialized Torch Models in Apache Spark (scala)
|
|
0
|
25
|
February 14, 2026
|
|
Pytorch dev 130
|
|
2
|
36
|
February 13, 2026
|
|
Is there any way to compute multiple jvps without repeatly computing function value?
|
|
5
|
44
|
February 13, 2026
|
|
Post training quantized model gets the error "Copying from quantized Tensor to non-quantized Tensor is not allowed" even though I'm not copying tensor
|
|
5
|
70
|
February 12, 2026
|
|
Mark_dynamic does not work when the dimension is 1
|
|
0
|
25
|
February 12, 2026
|
|
Is torch Muon optimizer compatible with FSDP/HSDP?
|
|
1
|
51
|
February 12, 2026
|
|
Flex_attention and torch.compile
|
|
1
|
45
|
February 12, 2026
|
|
Cuda Memory Use Batched Matrix Multiplication
|
|
7
|
2057
|
February 11, 2026
|
|
ONNX Community Survey 2026
|
|
0
|
28
|
February 11, 2026
|
|
Help training Titan+MIRAS, model learns to cheat loss
|
|
0
|
20
|
February 11, 2026
|
|
Is it safe to create new gpu tensor during cuda graph capturing?
|
|
3
|
53
|
February 10, 2026
|
|
Looking for matmul.h
|
|
1
|
39
|
February 10, 2026
|
|
"Offset increment outside graph capture encountered unexpectedly" is a specific PyTorch Windows issue that's known to occur with CUDA Graphs. This is a bug in PyTorch Windows builds
|
|
0
|
71
|
February 10, 2026
|
|
DEBUG=1 build with NDEBUG flag possible? (CFLAGS override related)
|
|
0
|
25
|
February 10, 2026
|
|
Reduce_sum() or sum operation (reduction operation) for floating point ops isnt accurate or not implementing kahan summation algorithm ,instead just returning a+b --->leading to vanishing gradient problem
|
|
1
|
34
|
February 10, 2026
|
|
Request: KV Cache Steering for VLM Hallucination Mitigation
|
|
0
|
23
|
February 9, 2026
|
|
DP with opacus: Want to understanding the functions
|
|
2
|
137
|
February 9, 2026
|