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.
 
 
 
 
 
 
yangruoqi713 7c3eed1b1d [MSLITE][Develop] optimize arm cpu fp16 op lstm and gru: use matmul fp16 5 years ago
..
activation_fp16.c [MSLITE] Add implementation for intReduceSum and fix bug of tanh 5 years ago
activation_fp16.h [MSLITE] arithmetic optimize 5 years ago
arg_min_max_fp16.c [MSLITE] argmin max fp16 5 years ago
arg_min_max_fp16.h [MSLITE] argmin max fp16 5 years ago
arithmetic_fp16.c add tensorlist and bias_add fp16 ops 5 years ago
arithmetic_fp16.h add tensorlist and bias_add fp16 ops 5 years ago
arithmetic_self_fp16.c [MSLITE] nnacl: concat cast expanddims reshape gather fill 5 years ago
arithmetic_self_fp16.h add reciprocal op 5 years ago
batchnorm_fp16.c fix codedex 5 years ago
batchnorm_fp16.h fix_batchnorm_output_fp16 5 years ago
cast_fp16.h [MSLITE] nnacl: concat cast expanddims reshape gather fill 5 years ago
common_func_fp16.c [MSLITE] deconv winograd fp16 neon 5 years ago
common_func_fp16.h [MSLITE] deconv winograd fp16 neon 5 years ago
conv_depthwise_fp16.c [MSLITE][Develop] fix bug of arm cpu deconv depthwise 5 years ago
conv_depthwise_fp16.h rename nnacl fp32 5 years ago
conv_fp16.c optimize fp16 winograd parallel execute 5 years ago
conv_fp16.h optimize fp16 common conv 5 years ago
crop_fp16.c fix code review 5 years ago
crop_fp16.h fix multipe definition of function 5 years ago
deconv_fp16.c fp16 deconv winograd 5 years ago
deconv_fp16.h fp16 deconv winograd 5 years ago
deconv_winograd_fp16.c [MSLITE] MAX_SHAPE_SIZE COMM_SHAPE_SIZE 5 years ago
deconv_winograd_fp16.h [MSLITE] deconv winograd fp16 neon 5 years ago
gru_fp16.c [MSLITE][Develop] optimize arm cpu fp16 op lstm and gru: use matmul fp16 5 years ago
gru_fp16.h [MSLITE][Develop] optimize arm cpu fp16 op lstm and gru: use matmul fp16 5 years ago
instance_norm_fp16.c [MSLITE][DEVELOP] add cpu fp16 op: instance norm 5 years ago
instance_norm_fp16.h [MSLITE][DEVELOP] add cpu fp16 op: instance norm 5 years ago
lstm_fp16.c [MSLITE][Develop] optimize arm cpu fp16 op lstm and gru: use matmul fp16 5 years ago
lstm_fp16.h [MSLITE][Develop] optimize arm cpu fp16 op lstm and gru: use matmul fp16 5 years ago
matmul_fp16.c [MSLITE] fp16 matmul 5 years ago
matmul_fp16.h [MSLITE] fp16 matmul 5 years ago
matrix_fp16.c [MSLITE] deconv winograd fp16 neon 5 years ago
matrix_fp16.h [MSLITE] deconv winograd fp16 neon 5 years ago
pack_fp16.c [MSLITE] arithmetic optimize 5 years ago
pack_fp16.h fp16 winograd init optimize 5 years ago
pad_fp16.c pad_fp16_support_reflect 5 years ago
pad_fp16.h rename nnacl fp32 5 years ago
pooling_fp16.c add activation fusion for fp16 pooling 5 years ago
pooling_fp16.h [MSLITE] arithmetic optimize 5 years ago
quant_dtype_cast_fp16.c quant cast fp16 support uint8 5 years ago
quant_dtype_cast_fp16.h quant cast fp16 support uint8 5 years ago
reduce_fp16.c [MSLITE][Develop] fix bug of arm fp16 cpu op: reduce, arithmetic 5 years ago
reduce_fp16.h reduce int8 malloc data assign data split 5 years ago
scale_fp16.c fix multipe definition of function 5 years ago
scale_fp16.h fix multipe definition of function 5 years ago
softmax_fp16.c Use local max input value instead of the global one for fp32 & fp16 softmax 5 years ago
softmax_fp16.h softmax activation fp16 5 years ago
transpose_fp16.c transpose remove unnecessary member 5 years ago
transpose_fp16.h transpose remove unnecessary member 5 years ago
winograd_transform_fp16.c fp16 deconv winograd 5 years ago
winograd_transform_fp16.h fp16 deconv winograd 5 years ago
winograd_utils_fp16.c add fp16 selection func 5 years ago
winograd_utils_fp16.h add fp16 selection func 5 years ago