Repository Issue Activity (beta)

#946 Proposed signature change for `tch::Tensor::grad`

#1026 feature request: `load-dynamic`

#1021 XPU backend(?)

#936 Do you expose torch.distributed or similar?

#1017 Need a new release so I can upgrade to torch 2.12.0

#1019 Improve Error Output for Simple Tensor Mismatch

#988 CUDA not available despite correct libTorch and Nvidia Driver

#1015 ROCm: the GPU backend gets dropped at link time (`--as-needed`), so `tch` reports CPU

#1012 add num-complex as an optional dependency?

#1011 Please allow for tch-rs discussions

#1009 Bump rand to 0.9.3+ to address RUSTSEC-2026-0097 unsoundness advisory

#1008 Varstore should be replace by a HashMap

#1007 varstore should support specifying dtype when create tensor.

#1006 Does some know how to use tch-rs on Fedora 43 ?

#234 Cuda Stream Support

#1004 loose the version constraint of ndarray?

#941 The argument 'device' of Tensor.pin_memory() is deprecated

#1002 Support for PyO3 0.25+ to enable Python 3.14 compatibility

#890 -

#937 -

#1001 Can I directly do object detection training and inference?

#998 Please update libtorch 2.9.1 version

#921 `torch_cuda.dll` not linked when using dynamic linking, and missing `clog.lib` when static linking

#997 Add `u16`, `u32`, and `u64` to `Kind`

#984 How to set the CUDA path?