87 Pull requests merged by 36 people

• 修复了paddle.unfold的大Tensor问题

  #74379 merged Aug 2, 2025

• 修复paddle.incubate.nn.functional.fused_rotary_position_embedding的非法地址访问问题

  #74347 merged Aug 2, 2025

• [User Experience] Remove nvprof_nvtx_pop() in gemm.py

  #74309 merged Aug 1, 2025

• disable test_custom_contiguous

  #74337 merged Aug 1, 2025

• synchronize before print bf16 cast

  #74278 merged Aug 1, 2025

• onednn/fc_kernel.cc add use_onednn [fluid_ops]

  #74325 merged Aug 1, 2025

• [0-size Tensor Retest] fix take_along_axis bug

  #74354 merged Aug 1, 2025

• [BIG tensor] fix overflow in cudnn_softmax by replacing the api

  #74329 merged Aug 1, 2025

• 【stride】 fix set_value stride_copy condition

  #74341 merged Aug 1, 2025

• [PIR save/load]Change pir_version to 0

  #74336 merged Aug 1, 2025

• paddle/fluid/operators/ops_extra_info.h add use_onednn [fluid_ops]

  #74327 merged Aug 1, 2025

• [BIG tensor] enhance robustness of softmax_gpudnn

  #74330 merged Aug 1, 2025

• [0-size Tensor No.237] Add 0-size Tensor support for paddle.nn.utils.vector_to_parameters

  #74348 merged Aug 1, 2025

• [0-size Tensor Retest] fix pool3d and adaptive_pool3d for 0-size

  #74333 merged Aug 1, 2025

• 【Inference Fix Bug】Fix stream_k in moe_gemm

  #74351 merged Aug 1, 2025

• [AutoParallel]Fix get_group method of processmesh

  #73099 merged Jul 31, 2025

• revert PR 74113

  #74339 merged Jul 31, 2025

• [bug fix] add default context parallel degree

  #74346 merged Jul 31, 2025

• [Stride]Fix stride permute bug

  #74312 merged Jul 31, 2025

• Replace is_run_mkldnn_kernel is_run_onednn_kernel [fluid_ops]

  #74314 merged Jul 31, 2025

• pp shared_comm

  #74113 merged Jul 31, 2025

• readd mkldnn api in paddle/fluid/inference/api/ [fluid_ops]

  #74230 merged Jul 31, 2025

• [Big Tensor]Fix accuracy diff for paddle.nn.functional.adaptive_max_pool3d API.

  #74279 merged Jul 31, 2025

• [0-size Tensor No.87、214]Add 0-size Tensor support for paddle.incubate.nn.functional. fused_rotary_position_embedding 、scaled_dot_product_attention

  #74323 merged Jul 31, 2025

• [Complex] Fix wrong reciprocal result when meets complex dtype

  #74290 merged Jul 31, 2025

• [AutoParallel] pp dataloader align mode

  #73941 merged Jul 31, 2025

• [0-size Tensor No.354、355] Add 0-size Tensor support for unique [fluid_ops]

  #74305 merged Jul 31, 2025

• 【Inference Optimize】Support setting environment variables to enable stream_k

  #74317 merged Jul 31, 2025

• [Accuracy diff No.90] Fix accuracy diff for paddle.incubate.nn.functional.fused_bias_dropout_residual_layer_norm API

  #74149 merged Jul 31, 2025

• Implement IEEE 754 rounding conditions for fp32 to fp16 conversion in host

  #74044 merged Jul 31, 2025

• fix cuda error 700 for softmax

  #74298 merged Jul 31, 2025

• Add check with cublas call gemmex

  #74306 merged Jul 31, 2025

• Fix einsum accuracy diff when with contraction dim

  #74303 merged Jul 31, 2025

• [Phi]Fix group_norm for big tensor

  #74172 merged Jul 31, 2025

• [CINN] Robustify min/max type matching

  #74316 merged Jul 31, 2025

• fix dual pp bug

  #74158 merged Jul 31, 2025

• Add paddle.device.memory and paddle.device.device_count API

  #74308 merged Jul 31, 2025

• support fa3 varlen bwd and modify fa3 varlen fwd api

  #73831 merged Jul 30, 2025

• [0-size Tensor Retest] add avg/max poll for float16

  #74313 merged Jul 30, 2025

• 【stride】update copy and set_value support stride input/output

  #74271 merged Jul 30, 2025

• [BIG tensor]fix overflow problem in addmm caused by the cublas api

  #74289 merged Jul 30, 2025

• [0-size Tensor Retest]Fix fold、fused_feedforward、lstsq、broadcast_tensor api to sopport 0-size Tensor

  #74295 merged Jul 30, 2025

• [BIG tensor]enhance robustness of mm ops

  #74293 merged Jul 30, 2025

• fix_linalg_solve

  #74282 merged Jul 30, 2025

• Replace cur_mkldnn_session_id cur_onednn_session_id [fluid_ops]

  #74244 merged Jul 30, 2025

• Revert "Fix paddle.vision.ops.deform_conv2d API big Tensor"

  #74304 merged Jul 30, 2025

• add is_compiled_with_onednn [fluid_ops]

  #74299 merged Jul 30, 2025

• [PHI]Fix nanmedian

  #74263 merged Jul 30, 2025

• [Typos] Rename namespace for TensorFormatter

  #74287 merged Jul 30, 2025

• [Typos] Fix paddle::operator::math to phi::math for depthwise_conv.h

  #74288 merged Jul 30, 2025

• [big tensor] upsample 支持大tensor

  #74242 merged Jul 30, 2025

• [Inference] Interleaving buffers in internode_ll_two_stage

  #74276 merged Jul 30, 2025

• support context parallel

  #74201 merged Jul 30, 2025

• [XPU] support fused_rms_norm_ext

  #74286 merged Jul 30, 2025

• Add not support enforce for lstsq big tensor

  #74280 merged Jul 30, 2025

• [distributed] fix bug when ep_degree == sharding_degree

  #74268 merged Jul 30, 2025

• [0-size Tensor No.82] Add 0-size Tensor support for paddle.incubate.nn.functional.fused_bias_act [fluid_ops]

  #74259 merged Jul 29, 2025

• [0-size Tensor No.358] Add 0-size Tensor support for paddle.vision.ops.generate_proposals [fluid_ops]

  #74261 merged Jul 29, 2025

• [Accuracy diff No.132] Fix accuracy diff for paddle.outer API

  #74182 merged Jul 29, 2025

• [Accuracy diff No.141] Fix accuracy diff for paddle.Tensor.__mul__ API

  #74198 merged Jul 29, 2025

• CINN support float8e4m3.

  #74111 merged Jul 29, 2025

• [Accuracy diff No.84-85] Fix accuracy diff for paddle.einsum API

  #74257 merged Jul 29, 2025

• [PHI]Fix slice empty value

  #74267 merged Jul 29, 2025

• [BIG tensor]Fix int32 overflow for l1_loss

  #74211 merged Jul 29, 2025

• [PHI] Move some core impl to .cc to avoid include glog in header

  #74281 merged Jul 29, 2025

• [XPU] reduce_min supports int64/int32

  #74275 merged Jul 29, 2025

• update block_radix_topk.cuh namespace

  #74232 merged Jul 29, 2025

• Rename context to dev_ctx in paddle/phi/kernels/funcs [fluid_ops]

  #74233 merged Jul 29, 2025

• Fix elemenewist_add_op elementwise_add_op [fluid_ops]

  #74245 merged Jul 29, 2025

• Async_offload does not support custom device, fall back to sync

  #74220 merged Jul 29, 2025

• [BIG tensor] Fix cuda error in addmm

  #74273 merged Jul 29, 2025

• Fix paddle.incubate.nn.functional.fused_bias_dropout_residual_layer_norm big Tensor

  #74060 merged Jul 29, 2025

• Reinforce send_ue_recv for big tensor

  #74213 merged Jul 29, 2025

• [PHI] Support paddle.linalg.lu in DCU

  #74258 merged Jul 29, 2025

• [BIG tensor] fix acc error in margin_cross_entropy

  #74252 merged Jul 29, 2025

• [PHI] Add missing const qualifier for Tensor.contiguous

  #74264 merged Jul 28, 2025

• [0-size Tensor No.353] Add 0-size Tensor support for unflatten API.

  #73986 merged Jul 28, 2025

• [CINN] Symplify gather_nd min/max type unifying

  #74256 merged Jul 28, 2025

• [Big-Tensor Fix] Add error throwing check of ctc_loss

  #74223 merged Jul 28, 2025

• [XPU] bump XRE to 5.0.21.26

  #74260 merged Jul 28, 2025

• [XPU] max_pool2d use xpudnn

  #74156 merged Jul 28, 2025

• [Distributed] Suooprt bubble hook for VPPFhenBInBalancedMemory

  #74099 merged Jul 28, 2025

• [Distributed] Suooprt bubble hook for VPPFhenBInBalancedMemory

  #74100 merged Jul 28, 2025

• [BIG tensor] fix acc error of margin f16

  #74254 merged Jul 28, 2025

• [0-size Tensor Retest]Fix max_pool3d、avg_pool3d、fused_bias_dropout_residual_layer_norm、fused_layer_norm、softmax_mask_fuse、conv3d

  #74229 merged Jul 28, 2025

• [0-size Tensor No.88] Add 0-size Tensor support for paddle.incubate.nn.functional.variable_length_memory_efficient_attention [fluid_ops]

  #74157 merged Jul 28, 2025

• [0-size Tensor Job2 No.51] Add 0-size Tensor support for paddle.multiplex [fluid_ops]

  #74212 merged Jul 28, 2025

60 Pull requests opened by 36 people

• [CI] slice no cache paddle

  #74265 opened Jul 28, 2025

• [CINN] support dcu get_value_in_kernel_args with sycl and hip

  #74266 opened Jul 28, 2025

• support w4afp8

  #74270 opened Jul 28, 2025

• [Deep_ep]Fix deepep calc_stream bug

  #74274 opened Jul 28, 2025

• [XPU] Update XHPC to 20250722.

  #74277 opened Jul 28, 2025

• Update deep_ep intranode & internode kernels

  #74284 opened Jul 29, 2025

• [PHI]Optimize Index_put and Set up Index Kernel Stride Mechanism

  #74291 opened Jul 29, 2025

• CINN Add more Simplify scenario: Select2MinMax, BoundSimplify, PowerL…

  #74292 opened Jul 29, 2025

• [CINN] dcu does not support fused op

  #74294 opened Jul 29, 2025

• Extend moe_permute & moe_unpermute for some incubating features

  #74296 opened Jul 29, 2025

• [Auto Parallel] Add spmd_rule about sharding on the same tensor dim by many mesh dim for softmax

  #74300 opened Jul 30, 2025

• Rpe attention

  #74310 opened Jul 30, 2025

• Add moe_permute & unpermute with do_gather(fill_x) feature to fleety

  #74311 opened Jul 30, 2025

• add FLAGS_use_onednn [fluid_ops]

  #74315 opened Jul 30, 2025

• 在scaled_dot_product_attention函数中，考虑mask是3D的情况

  #74319 opened Jul 30, 2025

• [Slice] Dispatch bool-only combined case to non-zero and elementwise_get

  #74320 opened Jul 30, 2025

• Test pr

  #74321 opened Jul 30, 2025

• correct copysign backward

  #74322 opened Jul 30, 2025

• Fix paddle.vision.ops.deform_conv2d API big Tensor

  #74324 opened Jul 30, 2025

• executor add use_onednn [fluid_ops]

  #74326 opened Jul 31, 2025

• [CINN] [New Hardware Update]:SYCL PR6: fix sycl bugs and sycl reuse paddle stream

  #74328 opened Jul 31, 2025

• [BIG tensor] 修复paddle.nansum/nanmean梯度计算

  #74332 opened Jul 31, 2025

• [big tensor] matmul cuda700问题

  #74335 opened Jul 31, 2025

• [CppExtension] Auto define `PADDLE_EXTENSION_NAME` macro

  #74338 opened Jul 31, 2025

• [BIG tensor] enhance robustness of bilinear

  #74342 opened Jul 31, 2025

• GradientClipByGlobalNorm与原来的动半pp逻辑对齐

  #74343 opened Jul 31, 2025

• [slice]add IndexPutWithSortKernel in index_elementwise_get_grad slice…

  #74344 opened Jul 31, 2025

• [CI] Fleety ci

  #74345 opened Jul 31, 2025

• add get_device_place in op_test.py, and replace the method of obtaining place in legacy_test

  #74349 opened Jul 31, 2025

• [Auto Parallel] Add spmd_rule about sharding on the same tensor dim by many mesh dim for reshape

  #74352 opened Jul 31, 2025

• test

  #74353 opened Jul 31, 2025

• caching links update [fluid_ops]

  #74356 opened Aug 1, 2025

• fix GetInputTypes() & GetOutputTypes() when using .json model

  #74358 opened Aug 1, 2025

• [big tensor] fix median hang

  #74360 opened Aug 1, 2025

• [CINN] [New Hardware Update]:Add CINN HIP fp16 support

  #74361 opened Aug 1, 2025

• 修复fused_feedforward的大tensor问题

  #74362 opened Aug 1, 2025

• split get_places function, add is_custom_device in op_test

  #74363 opened Aug 1, 2025

• Fix class center sample

  #74364 opened Aug 1, 2025

• CI测试不review analysis_config [fluid_ops]

  #74367 opened Aug 1, 2025

• [BIG tensor] add invalid shape check for preventing implicit overflow

  #74369 opened Aug 1, 2025

• [BIG tensor] enhance robustness of size calculation

  #74370 opened Aug 1, 2025

• [Custom Device] add float16 and bfloat16 supported api for custom device

  #74371 opened Aug 1, 2025

• Add LOG(WARNING) for kernel fallback

  #74372 opened Aug 1, 2025

• Fix set_paddle_lib_path

  #74373 opened Aug 1, 2025

• [Feat]: cpu max kernel support bool dtype

  #74374 opened Aug 1, 2025

• Flashmaskv2 fleety

  #74375 opened Aug 1, 2025

• Fix unused variable

  #74376 opened Aug 2, 2025

• 修复paddle.linalg.cholesky分解大Tensor的问题

  #74377 opened Aug 2, 2025

• 修复随机数某些平台生成重复数字

  #74378 opened Aug 2, 2025

• 修复了paddle.unique_consecutive的大Tensor问题

  #74380 opened Aug 2, 2025

• refine some error message to avoid linking words together part1

  #74381 opened Aug 3, 2025

• Fix interpolate

  #74382 opened Aug 3, 2025

• refine some error message to avoid linking words together part2

  #74383 opened Aug 3, 2025

• refine some error message to avoid linking words together part3

  #74384 opened Aug 3, 2025

• Fix typos gblic glibc

  #74385 opened Aug 3, 2025

• refine some error message to avoid linking words together part4

  #74386 opened Aug 3, 2025

• refine some error message to avoid linking words together part5

  #74387 opened Aug 3, 2025

• fix build scripts

  #74388 opened Aug 3, 2025

• [Feat]: add LibUVTCPServer backend for TCPStore

  #74389 opened Aug 3, 2025

• [0-size Retest] Fix outer for 0-size Tensor

  #74390 opened Aug 3, 2025

27 Issues closed by 11 people

• paddlex3.0版本依赖的numpy/pandas为什么要限制比较低的版本？

  #73068 closed Aug 1, 2025

• 【开源任务】 Paddle API 0-size 机制建设

  #72637 closed Aug 1, 2025

• 建议保持C++预测引擎ClearIntermediateTensor方法的向下兼容性

  #74004 closed Aug 1, 2025

• 【Hackathon 8th Overview】第八期飞桨黑客松活动总览

  #71315 closed Aug 1, 2025

• 【HACKATHON 8th Code Camp】飞桨正式实习招聘（可在校）

  #71313 closed Aug 1, 2025

• 【Hackathon 8th】开源贡献个人挑战赛

  #71310 closed Aug 1, 2025

• 【Hackathon 8th】Fundable Projects（偷跑版）

  #70860 closed Aug 1, 2025

• `paddle.reciprocal` returns NaN for infinities instead of zero on both CPU and GPU

  #73714 closed Jul 31, 2025

• paddlepaddle3.0.0适配的pytorch版本

  #74250 closed Jul 29, 2025

• nanmedian最近被改出bug，计算结果完全不对

  #74239 closed Jul 29, 2025

• thrid party里的flash-attn源码编译报错

  #73928 closed Jul 29, 2025

• Paddle 在rocm软件栈中的编译错误

  #66368 closed Jul 29, 2025

• PIR 单测推全交流平台

  #66134 closed Jul 29, 2025

• 在windows容器中使用paddlepaddle-gpu有问题

  #66112 closed Jul 29, 2025

• DeepXDE在paddle backend下使用L-BFGS优化器出现错误结果

  #66100 closed Jul 29, 2025

• 同样docker容器，同样的代码3090/4070显卡报错，V100/A100训练成功

  #66095 closed Jul 29, 2025

• 单卡训练报错ValueError: paddle.distributed initialize error, environment variable FLAGS_selected_gpus is needed, but not set.

  #66092 closed Jul 29, 2025

• 飞腾+麒麟v10+昆仑K100环境中编译 PaddlePaddle 报错

  #64903 closed Jul 29, 2025

• 静态图模式下通过nccl等方式完成流水线并行通信是否行得通

  #63934 closed Jul 29, 2025

• Abuses of logging.getLogger("root") harm unrelated innocent logger use. 滥用root logger影响所有logger的使用

  #63529 closed Jul 29, 2025

• arm架构安装paddle环境问题

  #55455 closed Jul 29, 2025

• GPU使用报错

  #74192 closed Jul 29, 2025

• inflateReset2 CRASH

  #70538 closed Jul 29, 2025

• 请问支持英伟达显卡5090吗？

  #73032 closed Jul 29, 2025

• 论文引用

  #69000 closed Jul 28, 2025

• 开源相关

  #69312 closed Jul 28, 2025

• 【开源任务】 Paddle API 0-size 机制建设-第2批

  #73182 closed Jul 28, 2025

9 Issues opened by 8 people

• GraphNet 共创计划

  #74368 opened Aug 1, 2025

• 3.x版本什么时候能够提供对昆仑芯 XPU 的 ARM 架构支持

  #74365 opened Aug 1, 2025

• CreateDiscountRequest does not support Discount Group ID

  #74355 opened Jul 31, 2025

• Paddle如何关闭检测GPU

  #74307 opened Jul 30, 2025

• docker 容器里安装的GPU版，在cpu环境下启动容器，执行预测会报错。

  #74302 opened Jul 30, 2025

• 依赖冲突

  #74301 opened Jul 30, 2025

• 【Hackathon 9th】护航计划集训营（提前批）

  #74285 opened Jul 29, 2025

• paddle找不到H20-3eGPU，无法训练

  #74283 opened Jul 29, 2025

• can not use SetModelBuffer

  #74272 opened Jul 28, 2025

81 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• CINN support float8 & Optimize in inline_tactic & SelectOp->min/max

  #72703 commented on Aug 2, 2025 ? 4 new comments

• [Auto-parallel] 增强stage对层间传递stop_grad=True的参数的支持

  #73459 commented on Aug 3, 2025 ? 4 new comments

• [XPU] support python streams api for xpu

  #73924 commented on Aug 1, 2025 ? 3 new comments

• Add custom device support for skipping unittest for non CUDA devices

  #74169 commented on Aug 2, 2025 ? 2 new comments

• [Auto Parallel]Add spmd_rule about sharding on the same tensor dim by many mesh dim for LayerNorm op and binary elementwise ops.

  #74253 commented on Aug 1, 2025 ? 2 new comments

• [BIG tensor]add PADDLE_ENFORCE_LE for unbind

  #74224 commented on Jul 31, 2025 ? 0 new comments

• [Fix] Add missing dependency

  #71651 commented on Aug 2, 2025 ? 0 new comments

• Paddle xpu chunk

  #71402 commented on Aug 2, 2025 ? 0 new comments

• Fix pnc bug

  #67802 commented on Aug 2, 2025 ? 0 new comments

• Test sd infer e2e

  #64330 commented on Aug 2, 2025 ? 0 new comments

• Test group norm dyshape

  #64162 commented on Aug 2, 2025 ? 0 new comments

• support 3d fused_rope

  #63526 commented on Aug 2, 2025 ? 0 new comments

• Ana modular ana bug

  #63440 commented on Aug 2, 2025 ? 0 new comments

• fix security concerns in topology.

  #63366 commented on Aug 2, 2025 ? 0 new comments

• [Inference cpu]add log_softmax cpu kernel AVX version

  #63331 commented on Aug 2, 2025 ? 0 new comments

• [Not Review] Fix the compiling with KP.

  #63316 commented on Aug 2, 2025 ? 0 new comments

• Disable mp_scale in RowSequenceParallelLinear because it is conflict with fused_linear_grad_add.

  #63149 commented on Aug 2, 2025 ? 0 new comments

• Add ConstraintsManager class

  #63110 commented on Aug 2, 2025 ? 0 new comments

• Add sycl and hip support into Paddle-CINN

  #63034 commented on Aug 2, 2025 ? 0 new comments

• [XPU] add useless pad op when padding all zero to identity_op pass;

  #62948 commented on Aug 2, 2025 ? 0 new comments

• 【Launch】Upgrade ease of launch

  #62920 commented on Aug 2, 2025 ? 0 new comments

• [CINN] Merge sycl hip work into develop

  #62888 commented on Aug 2, 2025 ? 0 new comments

• Add reshape expand constrain

  #62883 commented on Aug 2, 2025 ? 0 new comments

• [Fix Issue] fix issue 73692

  #74222 commented on Aug 4, 2025 ? 0 new comments

• update

  #74216 commented on Aug 2, 2025 ? 0 new comments

• Fix paddle.linalg.vector_norm for big tensor

  #74197 commented on Aug 2, 2025 ? 0 new comments

• [Big tensor] enhance fused linear?op

  #74194 commented on Aug 2, 2025 ? 0 new comments

• [Slice] Support multi-dim bool tensor index

  #74193 commented on Aug 2, 2025 ? 0 new comments

• [PHI] Fix `paddle.linalg.det` GradKernel using LU

  #74179 commented on Aug 1, 2025 ? 0 new comments

• [cherry-pick] [Distributed] fix eval batch & non-compute_loss in pipeline

  #74170 commented on Aug 2, 2025 ? 0 new comments

• refine arange check

  #74226 commented on Jul 30, 2025 ? 0 new comments

• [cherry-pick][Distributed] fix recreate nccl comm bug (#73625)

  #74168 commented on Aug 2, 2025 ? 0 new comments

• [XPU] set FLAGS_use_stride_kernel=1 for CI test

  #74161 commented on Aug 2, 2025 ? 0 new comments

• refine full and full_like for fill_value type check and annotations

  #74127 commented on Aug 2, 2025 ? 0 new comments

• pp shared_comm

  #74115 commented on Aug 2, 2025 ? 0 new comments

• [不 REVIEW] Slice benckmark CI

  #74074 commented on Jul 30, 2025 ? 0 new comments

• [CI] no fetch sot CINN

  #74035 commented on Jul 31, 2025 ? 0 new comments

• [AutoParallel]Change the trigger position of autodp fake r to s

  #74243 commented on Jul 31, 2025 ? 0 new comments

• [AutoParallel] Add features `_only_reshard_mesh_shape` and `get_local_slice`

  #74248 commented on Jul 28, 2025 ? 0 new comments

• [Accuracy diff No.42、118] Fix accuracy diff for paddle.nn.functional.embedding API

  #73445 commented on Jul 28, 2025 ? 0 new comments

• [Compat] POC for PyTorch C++ API compat conversion

  #72987 commented on Jul 29, 2025 ? 0 new comments

• Modify launch

  #59181 commented on Aug 2, 2025 ? 0 new comments

• [Paddle Dy2Static] Fix custom_op python api of inplace input in dy2static.

  #56776 commented on Aug 2, 2025 ? 0 new comments

• print op check info

  #55120 commented on Aug 2, 2025 ? 0 new comments

• Release/2.4

  #50898 commented on Aug 2, 2025 ? 0 new comments

• Ppyolo 2 4

  #50172 commented on Aug 2, 2025 ? 0 new comments

• checkout 10_9 paddle commit && complie with 2023 api

  #49545 commented on Aug 2, 2025 ? 0 new comments

• Merge gpugraph v2

  #48387 commented on Aug 2, 2025 ? 0 new comments

• Merge v2 develop

  #48363 commented on Aug 2, 2025 ? 0 new comments

• [Paddle-TRT][Cherry-Pick]Stride to 2 4

  #47135 commented on Aug 2, 2025 ? 0 new comments

• jit using paddle inference do mkldnn and mkldnn int8 infernce

  #46821 commented on Aug 2, 2025 ? 0 new comments

• [KP] simplification of compilation for xpu

  #41453 commented on Aug 2, 2025 ? 0 new comments

• dymf tmp

  #40497 commented on Aug 2, 2025 ? 0 new comments

• 【开源任务】Paddle CPU/GPU Kernel 精度问题推全

  #72667 commented on Aug 3, 2025 ? 0 new comments

• 飞桨支持5090 B架构编译，支持cmake4.0，支持3.8退场

  #73427 commented on Aug 1, 2025 ? 0 new comments

• paddle.distributed.rpc.init_rpc这个API无法运行，示例代码也不能跑

  #74246 commented on Jul 29, 2025 ? 0 new comments

• PPStructureV3预测时报错

  #74249 commented on Jul 29, 2025 ? 0 new comments

• nvidia-smi.exe检索GPU设备如何取消，默认使用cpu

  #74228 commented on Jul 29, 2025 ? 0 new comments

• paddlepaddle 3.0 save assign layer to model file failed

  #74206 commented on Jul 29, 2025 ? 0 new comments

• logsoftmax kernel 当输入中有 -inf 时在cpu和gpu上计算结果存在不一致

  #69859 commented on Jul 28, 2025 ? 0 new comments

• 有计划提供cuda 12.4、12.8的镜像或安装包吗？

  #72523 commented on Jul 28, 2025 ? 0 new comments

• Add nvtx tag for pp training.

  #62847 commented on Aug 2, 2025 ? 0 new comments

• Test llama infer e2e

  #62786 commented on Aug 2, 2025 ? 0 new comments

• Fuse gn silu for xpu

  #62685 commented on Aug 2, 2025 ? 0 new comments

• [CINN] Add constraints of symbol expr in flag

  #62620 commented on Aug 2, 2025 ? 0 new comments

• PP Support Send/Recv Dict[str, Tensor] between stages

  #62563 commented on Aug 2, 2025 ? 0 new comments

• [SOT] fix bug in llm BLIP

  #62555 commented on Aug 2, 2025 ? 0 new comments

• Test dy shape rms

  #61812 commented on Aug 2, 2025 ? 0 new comments

• Fix rms norm

  #61583 commented on Aug 2, 2025 ? 0 new comments

• Debug rms norm

  #61422 commented on Aug 2, 2025 ? 0 new comments

• [CINN] Combine Change of GatherNd, Slice, Concat and Try to Run It

  #61118 commented on Aug 2, 2025 ? 0 new comments

• [Paddle Inference] fix identity_op_clean_pass, add_support_int8_pass

  #60913 commented on Aug 2, 2025 ? 0 new comments

• [Paddle Inference] support TensorRT-LLM

  #60812 commented on Aug 2, 2025 ? 0 new comments

• [Paddle Inference] Fix blockmha bfloat16 compile error.

  #60558 commented on Aug 2, 2025 ? 0 new comments

• Fine grained 5 add attention mask

  #60311 commented on Aug 2, 2025 ? 0 new comments

• rename_skip_quant

  #60176 commented on Aug 2, 2025 ? 0 new comments

• Support gqa llama

  #59646 commented on Aug 2, 2025 ? 0 new comments

• Merge into develop part-3

  #59643 commented on Aug 2, 2025 ? 0 new comments

• feat(custom_reducer): size_1_tensor_not_concat/split

  #59490 commented on Aug 2, 2025 ? 0 new comments

• [Paddle Inference] Fix custom op exporting api error

  #59484 commented on Aug 2, 2025 ? 0 new comments

• [Paddle Inference] Fix mmha when qkv input is gqa.

  #59319 commented on Aug 2, 2025 ? 0 new comments