| .. |
|
distribution
|
GPU supports p2p nccl interfaces
|
2020-11-24 14:17:53 +08:00 |
|
mpi
|
code warning clean
|
2020-09-21 09:44:39 +08:00 |
|
blocking_queue.cc
|
add push opt logic
|
2020-12-15 15:35:42 +08:00 |
|
blocking_queue.h
|
add input data type check for ps cache mode
|
2021-02-07 12:35:04 +08:00 |
|
cuda_common.h
|
add more dtypes support for gatherdgrad and other bugfix
|
2020-11-30 14:00:54 +08:00 |
|
cuda_driver.cc
|
fix an allreduce bug with two streams sync problem
|
2021-05-10 10:00:58 +08:00 |
|
cuda_driver.h
|
fix an allreduce bug with two streams sync problem
|
2021-05-10 10:00:58 +08:00 |
|
cuda_env_checker.cc
|
search nvcc in entire PATH
|
2020-09-04 15:55:48 +08:00 |
|
cuda_env_checker.h
|
search nvcc in entire PATH
|
2020-09-04 15:55:48 +08:00 |
|
gpu_bucket.cc
|
add op atomic clean to clear input addr in launch allreduce
|
2021-03-25 11:19:23 +08:00 |
|
gpu_bucket.h
|
add op atomic clean to clear input addr in launch allreduce
|
2021-03-25 11:19:23 +08:00 |
|
gpu_buffer_mgr.cc
|
add error log when set device id failed
|
2021-03-25 14:58:03 +08:00 |
|
gpu_buffer_mgr.h
|
add push opt logic
|
2020-12-15 15:35:42 +08:00 |
|
gpu_common.h
|
addtensor size limitation to 2G
|
2021-04-06 20:29:31 +08:00 |
|
gpu_device_address.cc
|
use host_type vs deprecated type_id
|
2021-04-26 16:56:17 -04:00 |
|
gpu_device_address.h
|
fix device memory leak
|
2021-01-08 16:02:54 +08:00 |
|
gpu_device_manager.cc
|
fix an allreduce bug with two streams sync problem
|
2021-05-10 10:00:58 +08:00 |
|
gpu_device_manager.h
|
fix an allreduce bug with two streams sync problem
|
2021-05-10 10:00:58 +08:00 |
|
gpu_event.cc
|
PyNative AllReduce Bucket
|
2021-03-03 15:47:18 +08:00 |
|
gpu_event.h
|
PyNative AllReduce Bucket
|
2021-03-03 15:47:18 +08:00 |
|
gpu_kernel_build.cc
|
move the akg kernel build timer into AkgKernelBuilder::AkgKernelParallelBuild, so that it can time the Ascend kernel builder
|
2021-05-06 10:02:25 +08:00 |
|
gpu_kernel_build.h
|
add hardware abstract layer
|
2021-03-19 11:28:53 +08:00 |
|
gpu_kernel_runtime.cc
|
fix an allreduce bug with two streams sync problem
|
2021-05-10 10:00:58 +08:00 |
|
gpu_kernel_runtime.h
|
fix an allreduce bug with two streams sync problem
|
2021-05-10 10:00:58 +08:00 |
|
gpu_launch_kernel.cc
|
add hardware abstract layer
|
2021-03-19 11:28:53 +08:00 |
|
gpu_launch_kernel.h
|
add op atomic clean to clear input addr in launch allreduce
|
2021-03-25 11:19:23 +08:00 |
|
gpu_launch_mul.cc
|
add op_mul fusion based on allreduce fusion in pynative mode
|
2021-03-15 21:12:14 +08:00 |
|
gpu_launch_mul.h
|
add op atomic clean to clear input addr in launch allreduce
|
2021-03-25 11:19:23 +08:00 |
|
gpu_memory_allocator.cc
|
Refactor ms_context implementation
|
2020-08-30 22:51:40 +08:00 |
|
gpu_memory_allocator.h
|
Unified code style
|
2020-07-20 10:51:55 +08:00 |
|
gpu_memory_copy_manager.cc
|
fix an allreduce bug with two streams sync problem
|
2021-05-10 10:00:58 +08:00 |
|
gpu_memory_copy_manager.h
|
fix an allreduce bug with two streams sync problem
|
2021-05-10 10:00:58 +08:00 |
|
gpu_memory_manager.cc
|
add the continue memory alloc of communication kernel for actor runtime
|
2021-04-26 10:30:25 +08:00 |
|
gpu_memory_manager.h
|
profiler memory
|
2021-01-19 10:21:43 +08:00 |
|
gpu_stream_assign.cc
|
fix an allreduce bug with two streams sync problem
|
2021-05-10 10:00:58 +08:00 |
|
gpu_stream_assign.h
|
fix_consecutive_allreduce_bug
|
2020-08-20 16:04:56 +08:00 |
|
kernel_info_setter.cc
|
IR operators of GPU and CPU are unified as batchnorm
|
2021-03-18 19:02:28 +08:00 |
|
kernel_info_setter.h
|
fix graph output address set in the one time memory application scenarios
|
2021-04-26 16:02:18 +08:00 |
|
queue_common.h
|
add trace for gpu error/excpt log
|
2020-12-04 14:03:36 +08:00 |
|
readme.md
|
mindspore path adjust
|
2020-07-14 18:07:28 +08:00 |
|
trt_loader.cc
|
tensor-rt library dynamic loadg
|
2021-03-26 11:29:47 +08:00 |
|
trt_loader.h
|
gpu inference mixed precision
|
2021-04-25 09:53:47 +08:00 |
mindspore-ci-bot
84607e3a51