COMMITS
May 15, 2026
S
fp8_quant_blockwise supports very big tensor (#78977)
Shuhao Liang committed
May 14, 2026
X
[API compatibility] Align torch.nn.Module (#78836)
Xiaochun Yang committed
W
[ROCm 7.0] Add support for AMD CDNA4 and ROCm 7.0 (#77641)
WILSON WEI committed
Z
[Bug] fix view infershape bug (#78984)
zhwesky2010 committed
Z
fix(baddbmm_grad): use batched GEMM, gate split-scale behind flag (#78970)
Zhaowu Pan committed
Z
fix(addmm): preserve float32 precision for alpha/beta in bf16/fp16 GEMM (#78960)
Zhaowu Pan committed
X
[bugfix] fix shared_layer in MuonShardingOptimizer (#78975)
XiangzheWang committed
F
[XPU] Support partial rotary embedding in fused rope (#78976)
Fang Ru committed
May 13, 2026
N
[Compat] Force create arange attributes on CPU (#78981)
Nyakku Shigure committed
G
[CI] Update paddlefleet-ops install method (#78983)
Gu Shiwei committed
S
fix permute big tensor error (#78973)
SUN Dong committed
Z
fix fastdeploy compile bug (#78969)
Zhenghai Zhang committed
L
20260416 add ai edited test (#78690)
liuhao2638 committed
May 12, 2026
G
[CI] fix fleet build with python version (#78962)
Gu Shiwei committed
W
[API Compatibility] add pin_memory for randint (#78823)
Wenfei (Charles) Qi committed
A
[API Compatibility] Add alias for optimizers in `paddle.optimizer` (#78931)
ALGO1832 committed
X
[API compatibility] Align torch.inference_mode (#78905)
Xiaochun Yang committed
G
fix (#78947)
Gu Shiwei committed
Z
Fix learning_rate precision in SGD GPU kernel to use MT instead of T (#78761)
zhanghonggeng committed
May 11, 2026
S
optimize muon distributed optimizer (#78894)
ShenLiang committed
Z
[Performance] Optimize AdamW GPU kernel and Change the learning rate type to float64 (#78830)
zhengshengning committed
W
[API Compatibility] add paddle.cuda.OutOfMemoryError -part (#78874)
Wenfei (Charles) Qi committed
A
Z
S
[FlexCheckPoint] fix memory leaking of a recursive function (#78922)
Shuhao Liang committed
May 9, 2026
G
update (#78927)
Gu Shiwei committed
X
Add deepseekv4 Newton-Schulz coefficient set and update test coverage (#78916)
Xiangrui Yu committed
N
[Compat] Use public compat APIs for proxy controls (#78923)
Nyakku Shigure committed
L
fix recompute detection bug (#78911)
liufengwei0103 committed
W
[API Compatibility] add attribute `itemsize` to paddle.dtype -part (#78897)
Wenfei (Charles) Qi committed