Search results

torch sdpa returns nan output

Status: Open.

#1920 In intel/torch-xpu-ops;

· kaixuanliu opened on Aug 8, 2025

[TorchAO] AWQ path fails after PR2400 with `ImportError: cannot import name 'awq_uintx' from 'torchao.prototype.awq'

module: ao

Status: Open.

Bug

#1919 In intel/torch-xpu-ops;

· MingxuZh opened on Aug 8, 2025

[TorchAO] <code>_convert_weight_to_int4pack</code> XPU has different return shape with CUDA

module: ao

module: op impl

Status: Open.

Feature

#1914 In intel/torch-xpu-ops;

· Stonepia opened on Aug 7, 2025

[TorchAO] Int4XPULayout Lacks MoE Quant Support

dtype: int4

module: ao

module: quant

Status: Open.

Feature

#1913 In intel/torch-xpu-ops;

· Stonepia opened on Aug 7, 2025

· 

PT2.10

Implement the torch.ops.aten._weight_int4pack_mm for dequantizing the CUDA int4 layout

Status: Open.

#1912 In intel/torch-xpu-ops;

· yuanwu2017 opened on Aug 7, 2025

[Linux][BMG] bitsandbytes LLama3.1-8B INT8 finetune gets OOM

Status: Open.

#1911 In intel/torch-xpu-ops;

· XumingGai opened on Aug 7, 2025

with XPU build got "RuntimeError: MKL FFT error: Intel oneMKL DFTI ERROR: Inconsistent configuration parameters" in test_variant_consistency_eager cpu tests

module: ut

Status: Open.

Bug

#1910 In intel/torch-xpu-ops;

· daisyden opened on Aug 6, 2025

· 

PT2.9

implement fused rms norm

kernel_optimization

performance

Status: Open.

Feature

#1905 In intel/torch-xpu-ops;

· jianyizh opened on Aug 5, 2025

· 

PT2.10

[ARL-Windows] 97 UT Cases have torch.autograd.gradcheck.GradcheckError: Jacobian mismatch for output 0 with respect to input 1

bug_fix_stage2

Status: Open.

#1904 In intel/torch-xpu-ops;

· libohao1201 opened on Aug 4, 2025

implement torch.linalg.pinv xpu backend

module: op impl

Status: Open.

Feature

#1902 In intel/torch-xpu-ops;

· yao-matrix opened on Aug 1, 2025

· 

PT2.9

implement torch.linalg.svd xpu backend

module: op impl

Status: Open.

Feature

#1901 In intel/torch-xpu-ops;

· yao-matrix opened on Aug 1, 2025

· 

PT2.10

implement torch.linalg.qr XPU backend

module: op impl

Status: Open.

Feature

#1900 In intel/torch-xpu-ops;

· yao-matrix opened on Aug 1, 2025

· 

PT2.9