You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 
 
 
i-robot 96f56b827e !27203 while inplace op for cpu backend 4 years ago
..
arrays while inplace op for cpu backend 4 years ago
control code check for ops 4 years ago
cuda_impl rl discounted return 4 years ago
custom apply autodiff in custom 4 years ago
data support start profiler in the minddle of training. 4 years ago
debug add validation of vector size and non-zero validation of denominator for nn gpu operators. 4 years ago
math optimize error about BroadcastOpGrad 4 years ago
nccl Adapt nccl gpu kernel for compatibility. 4 years ago
nn optimizes the kernel error description of LSTM, Pad, ReLU, etc. 4 years ago
other change test level 4 years ago
quant clean up the static alarms of the second batch of operator groups 4 years ago
random clean up the static alarms of the second batch of operator groups 4 years ago
rl !27219 rl discounted return 4 years ago
sponge fix ops 4 years ago
trt add exception level to python error report 4 years ago
gpu_kernel.cc fix issue#I3ARG6 5 years ago
gpu_kernel.h add vector size check, nullptr check and clean code for gpu operators. 4 years ago
gpu_kernel_factory.cc refine log of kernel select 4 years ago
gpu_kernel_factory.h clean up the static alarms of the second batch of operator groups 4 years ago
kernel_constants.h add cholesky, cho_factor primitive and backend gpu implements 4 years ago