Repository Issue Activity (beta)

#79670 [Performance] cummax/cummin 的 CUDA kernel 硬编码 dim3(16,32)，长 innermost 维扫描比 torch.cummax 慢 19 倍、比 paddle.cumsum 慢 87 倍

#77340 NotImplementedError in oneDNN PIR conversion breaks CPU inference in PaddlePaddle 3.3.0

#79384 Paddle CI C++20 工具链升级清单

#76301 【启航计划】PaddlePaddle API兼容性增强

#79706 Conv2D 默认初始化(`Normal(0, √(2/fan_in))`)在无归一化深层/共享权重结构下激活指数放大，导致训练发散与 NaN 梯度

#79701 容器内 GPU MMU Fault: ENGINE GRAPHICS GPC1 GPCCLIENT_T1_3，导致推理回退到 CPU，重启容器后问题解决

#74692 Build fails: ModuleNotFoundError: No module named 'env_dict' (python/env_dict.py.in wasn't converted to python/env_dict.py)

#73078 Paddle3.0引用的Onednn（mkldnn）能否降到3.1.1版本

#74515 无法安装飞桨2.2版本

#74476 如何提高推理响应速度,现在的平均响应速度是143秒。8张96GB的昆仑芯P800，响应速度还不如4张英伟达A6000+vllm的响应速度74秒

#79618 paddle.i0 prematurely overflows to Inf for finite float64 result

#79436 Enable CI Support for ppc64le (IBM Power) Architecture

#73990 910B环境安装使用Paddle报NpuAclnnOpRunner: aclnnInplaceFillScalar

#72974 paddlepaddle 3.0.0可以在centos 7上使用吗

#79637 在windows上为rocm编译发生错误 Paddle only supports Python version >= 3.10 now

#79630 paddle.nn.functional.silu backward returns -Inf for finite float64 gradient at x=-709

#79629 paddle.nn.functional.selu backward returns wrong finite gradient at x=-700

#79628 paddle.nn.functional.elu backward returns 0.0 for finite float64 gradient at x=-40

#79627 paddle.expm1 backward returns 0.0 for finite float64 gradient at x=-40

#79527 Support Python 3.14 for PaddlePaddle

#79626 When will PaddlePaddle be available/kompatible for Python v3.14?

#79621 paddle.slice_scatter segfaults instead of raising for length-mismatched control args

#79620 paddle.hypot overflows to Inf for large finite float64 inputs

#79619 paddle.i1 prematurely overflows to Inf for finite float64 result

#79611 [Bug] PD_TensorSetStrides truncates int64 strides to int on macOS x64 (c_tensor.cc:272)