1. 01 9月, 2021 8 次提交
  2. 31 8月, 2021 8 次提交
  3. 30 8月, 2021 7 次提交
  4. 27 8月, 2021 9 次提交
  5. 26 8月, 2021 8 次提交
    • J
      [oneDNN] disable caching oneDNN primitives in matmul v2, Reduce grad and... · 31f0221f
      Jacek Czaja 提交于
      [oneDNN] disable caching oneDNN primitives in  matmul v2, Reduce grad and elementwise_add grad, expand_v2 (#35132)
      
      * - grad caching disabled of matmul_v1
      
      - compilation fix
      
      - compilation fix
      
      * - reduction removed
      
      * - Matmul v2 disabled caching
      
      * Draft of further changes
      
      * - workaround for reducegrad
      
      * - fixes to UT
      
      * - fix to compilation
      
      * - another fix
      
      * - fix
      31f0221f
    • S
      Add paddle.utils.dlpack APIs (#35067) · 8dc050d8
      Siming Dai 提交于
      * add dlpack api and fix a from_dlpack 
      8dc050d8
    • D
      fix assign bug support fp16 uint8 (#35153) · 270efb96
      duanboqiang 提交于
      * fix assign bug support fp16 uint8
      
      * fix dygragh assign bool bug
      
      * modify code style
      
      * revoke bool modification
      270efb96
    • W
      gc for newexecutor (#35085) · f1472039
      wanghuancoder 提交于
      * gc for newexecutor, test=develop
      
      * refine, test=develop
      
      * add interpretercore_gc_helper.h,test=develop
      
      * backup
      
      * gc whit thread and device_event, test=develop
      
      * refine, test=develop
      
      * refine, test=develop
      
      * refine, test=develop
      
      * refine, test=develop
      
      * fix bug, test=develop
      
      * refine, test=develop
      
      * refine, test=develop
      
      * refine, test=develop
      
      * add CheckGC, test=develop
      f1472039
    • S
      Support dropout backward in eval mode (#35122) · f1275fb6
      smallv0221 提交于
      * Support dropout backward in eval mode
      
      * add downscale case
      
      * minor fix
      
      * minor fix
      f1275fb6
    • W
      support tensor index. (#34824) · e7df47ec
      WeiXin 提交于
      * polish code
      
      * polish code.
      
      * polish code.
      
      * polish code.
      
      * polish code.
      e7df47ec
    • A
      Support Multi-Stream, Single-Thread in New Executor (#35024) · 678a259a
      Aurelius84 提交于
      * Modify into QueueSync QueueAsync
      
      * fix complie on MacOS
      
      * fix pointer
      
      * fix conflict
      
      * polish unittest
      
      * fix windows fetch error
      
      * polish code according reviewer
      
      * fix device_guard on CPU place
      678a259a
    • L
      Add feed_forward for fused attention op. (#34945) · d1a33bc7
      Li Min 提交于
      Describe
      
      Add feed_forward for fused attention op.
      (1) Encapsulate matmul impl (forward and backward) used in attention op.
      (2) Implement bias_add (forward and backward) used in attention op.
      d1a33bc7