Paddle icon indicating copy to clipboard operation
Paddle copied to clipboard

PArallel Distributed Deep LEarning: Machine Learning Framework from Industrial Practice (『飞桨』核心框架,深度学习&机器学习高性能单机、分布式训练和跨平台部署)

Results 1049 Paddle issues
Sort by recently updated
recently updated
newest added
trafficstars

### PR types Others ### PR changes OPs ### Describe Add group norm trt plugin support for dynamic and static shape, with coverage probe

### PR types Others ### PR changes Others ### Describe Onnx 需要WhileOp的输入和输出满足特定关系: 输出必定在输入中,因此修改了while的输入输出解析逻辑,满足了上述的要求。同时也需要适配将 CreateUndefinedVariable 函数中的 assign 操作放到 block0 中,否则 whileop 会在每个Step TensorCopy的时候发现输入没有被初始化。 ``` for i in range(tensor): s += i...

### PR types Others ### PR changes Others ### Describe refact fleet_moe.

### PR types New features ### PR changes APIs ### Describe instead of save ptq model, then load it before qat, now we can initialize qat scale directly with ptq...

### PR types Function optimization ### PR changes OPs ### Describe make affine_grid_op support 5d input_dim on cpu and gpu

### PR types Others ### PR changes Others ### Describe add macro PADDLE_WITH_XPU_BKCL to enforce_xpu.h

### PR types Bug fixes ### PR changes Others ### Describe Fix unqie_kernel cuda write out of bounds bug

### PR types Others ### PR changes Others ### Describe 1. 为新动态图的op收集input shapes 2. 加入接口的控制开关,控制是否收集input shapes和memory信息

### PR types Others ### PR changes Others ### Describe decouple Fleet with Optimizer、Model、Scaler.

### PR types New features ### PR changes OPs ### Describe elementwise pow raw support float16

PaddlePaddle Hackathon
contributor