tracel-ai/burn

Issues

• Tensor::slice incorrect beavior

  #3794 opened a month ago by Iskander-31
  2

• VecMat fusion autotune bug

  #3683 opened a month ago by laggui
  1

• Add torch.cross

  #3732 opened a month ago by torsteingrindvik
  0

• Unit tests are failing on Mac

  #3780 opened a month ago by antimora
  2

• vulkan backend - no possible adapter (Alpine)

  #3779 opened a month ago by andar1an
  9

• Mac M4 device=metal crash

  #3789 opened a month ago by OstrichRunning
  0

• Support referencing static memory for tensor data in embedded/WASM targets

  #3753 opened 2 months ago by antimora
  7

• Slice op with arbitraty steps

  #3690 opened a month ago by notdanilo
  4

• Implement optimized `fmod` operation for each backend, including burn-autodiff

  #3769 opened 2 months ago by antimora
  0

• Implement optimized `trunc` operation for each backend, including burn-autodiff

  #3768 opened 2 months ago by antimora
  0

• Module cloning is inefficient for CPU devices

  #3754 opened 2 months ago by antimora
  0

• broken tensor bitwise right shift with signed int on vulkan

  #3735 opened 2 months ago by tzemanovic
  3

• Gather in onnx-ir can't handle negative indices

  #3739 opened 2 months ago by vesuvisian
  0

• Failing to import ONNX with error: "No variable with name".

  #3650 opened 2 months ago by ArtificialCactus
  1

• Relax Element Trait constraints

  #3712 opened 2 months ago by skewballfox
  2

• `squeeze_dims` error on a imported ONNX model

  #3734 opened 2 months ago by A2va
  2

• SafetensorsFileRecorder: Memory‑efficient streaming load

  #3598 opened 2 months ago by antimora
  0

• NaN Gradients When Using mask_where with False Mask on Tensors Containing inf Values

  #3727 opened 2 months ago by L-M-Sherlock
  3

• Implement optimized bool select for each backend

  #3697 opened 2 months ago by antimora
  0

• Fusion backend limits size of multiplication scalars

  #3659 opened 2 months ago by hackaugusto
  3

• Clarify `init_lr` in NoamLrSchedulerConfig (acts as scaling factor, not initial starting LR)

  #3665 opened 2 months ago by Glimmering-Stallion
  1

• RFC: Burn‑audio crate — STFT/ISTFT, windows, overlap‑add, mel

  #3725 opened 2 months ago by ariawisp
  1

• Conv Ouput Geometry issues

  #3705 opened 2 months ago by crutcher
  5

• module 'pyarrow' has no attribute 'PyExtensionType'

  #3699 opened 2 months ago by cetra3
  3

• [Metal + Fusion] Creating a few BatchNorm layers panics with "out of memory" or "Ordering is bigger than operations"

  #3667 opened 2 months ago by AdrianEddy
  4

• Tons of onnx tests are failing when enabled for metal backend

  #3635 opened 2 months ago by antimora
  3

• Shouldn't `burn-cpu` have `dependencies.burn-tensor.default-features = false`?

  #3674 opened 2 months ago by dcrewi
  0

• burn-tch tests are failing on Mac

  #3676 opened 2 months ago by antimora
  3

• i64 tensor matmul memory validation

  #3660 opened 2 months ago by tzemanovic
  1

• Boolean mask indexing

  #3669 opened 2 months ago by laggui
  0

• Model does not seem to learn?

  #3661 opened 2 months ago by AspadaX
  0

• Interpolate disregards input dtype

  #3633 opened 2 months ago by torsteingrindvik
  2

• unresolved import `cubecl::cpu::CpuDevice`

  #3655 opened 2 months ago by buttfa
  1

• Random seed doesn't probagate correctly

  #3639 opened 2 months ago by nathanielsimard
  0

• Default tensor dtype can be configured on the device instead of by a type.

  #3642 opened 2 months ago by nathanielsimard
  3

• Dynamic Rank Tensor

  #3624 opened 2 months ago by nathanielsimard
  12

• ONNX import of Depth Anything v2

  #3640 opened 2 months ago by jpteb
  2

• Tensors: Allow choosing dtype on creation

  #3632 opened 2 months ago by torsteingrindvik
  1

• Memory Hint: Improved Memory Management

  #3638 opened 2 months ago by nathanielsimard
  0

• Reshape, permute, then full slice, breaks Tensor strides on WGPU

  #3629 opened 2 months ago by TomWyllie
  1

• 33 out of 259 tests are failing for Candle backend when testing onnx-tests

  #3634 opened 2 months ago by antimora
  0

• Default backend implementation

  #3628 opened 2 months ago by nathanielsimard
  2

• Update book to show how to integrate new kernels with the fusion backend.

  #3622 opened 2 months ago by hackaugusto
  0

• Implement memory-efficient ONNX weight loading (lazy/protobuf streaming)

  #3596 opened 3 months ago by antimora
  2

• `PyTorchFileRecorder` is unable to import snapshots from older pickle protocol versions

  #3605 opened 3 months ago by dcrewi
  1

• YOLO11x model output does not match reference on ndarray and metal backends

  #3600 opened 3 months ago by antimora
  6

• relu (at least for NdArray) has stride-related bug

  #3602 opened 2 months ago by crutcher
  1

• TUI Renderer is broken

  #3607 opened 2 months ago by siennathesane
  1

• PyTorchFileRecorder: Memory‑efficient streaming load

  #3597 opened 3 months ago by antimora
  0

• Linking errors on Win 10 & Nvidia RTX 5090

  #3591 opened 3 months ago by AndrejMitrovic
  1