InfiniTensor/src/operators
xiaonans a98573990b
Accelerate llama (#219)
* [feature] add cudagraph support

* modify code to pass the cuda_all_reduce test

* modify rope op

* support rmsnorm

* add fp16 support to silu cuda op

* fix bugs in rmsnorm op

* uncomment simplify in onnx.py

---------

Co-authored-by: Haojie Wang <haojie0429@gmail.com>
2024-04-01 08:46:05 +08:00
..
G2BMM.cc support Dynamic tensor infer shape and fix memory pool (#176) 2023-11-23 13:11:50 +08:00
GBMM.cc support Dynamic tensor infer shape and fix memory pool (#176) 2023-11-23 13:11:50 +08:00
activation_backward.cc support Dynamic tensor infer shape and fix memory pool (#176) 2023-11-23 13:11:50 +08:00
all_gather.cc support Dynamic tensor infer shape and fix memory pool (#176) 2023-11-23 13:11:50 +08:00
all_reduce.cc impl distributed launch with NCCL (#106) 2023-09-05 09:47:35 +08:00
attention_kvcache.cc use workspace to optimize kvcache attention 2024-01-25 10:33:01 +08:00
batch_norm.cc support Dynamic tensor infer shape and fix memory pool (#176) 2023-11-23 13:11:50 +08:00
broadcast.cc impl distributed launch with NCCL (#106) 2023-09-05 09:47:35 +08:00
concat.cc support Dynamic tensor infer shape and fix memory pool (#176) 2023-11-23 13:11:50 +08:00
conv.cc support Dynamic tensor infer shape and fix memory pool (#176) 2023-11-23 13:11:50 +08:00
det.cc support Dynamic tensor infer shape and fix memory pool (#176) 2023-11-23 13:11:50 +08:00
dropout.cc support Dynamic tensor infer shape and fix memory pool (#176) 2023-11-23 13:11:50 +08:00
element_wise.cc support Dynamic tensor infer shape and fix memory pool (#176) 2023-11-23 13:11:50 +08:00
expand.cc support Dynamic tensor infer shape and fix memory pool (#176) 2023-11-23 13:11:50 +08:00
extend.cc support Dynamic tensor infer shape and fix memory pool (#176) 2023-11-23 13:11:50 +08:00
gather.cc support Dynamic tensor infer shape and fix memory pool (#176) 2023-11-23 13:11:50 +08:00
gather_elements.cc support Dynamic tensor infer shape and fix memory pool (#176) 2023-11-23 13:11:50 +08:00
layer_norm.cc Modify kernel registration & support fp16 (#205) 2024-01-15 11:02:13 +08:00
lrn.cc Fix bang (#198) 2023-12-28 13:44:10 +08:00
matmul.cc feature: add parameter to config matmul compute type (#218) 2024-03-26 09:00:45 +08:00
membound.cc support Dynamic tensor infer shape and fix memory pool (#176) 2023-11-23 13:11:50 +08:00
pad.cc support Dynamic tensor infer shape and fix memory pool (#176) 2023-11-23 13:11:50 +08:00
pooling.cc support Dynamic tensor infer shape and fix memory pool (#176) 2023-11-23 13:11:50 +08:00
recv.cc Add send and recv operators based on NCCL (#182) 2023-12-14 16:38:03 +08:00
reduce.cc Add ReduceSum op and kernel (#160) 2023-11-24 09:29:58 +08:00
reshape.cc support Dynamic tensor infer shape and fix memory pool (#176) 2023-11-23 13:11:50 +08:00
resize.cc support Dynamic tensor infer shape and fix memory pool (#176) 2023-11-23 13:11:50 +08:00
rms_norm.cc Accelerate llama (#219) 2024-04-01 08:46:05 +08:00
rope.cc add test for rotary embedding cuda kernel 2024-02-04 10:24:20 +08:00
send.cc Add send and recv operators based on NCCL (#182) 2023-12-14 16:38:03 +08:00
slice.cc support Dynamic tensor infer shape and fix memory pool (#176) 2023-11-23 13:11:50 +08:00
softmax.cc support mixed dtype (#102) 2023-08-16 21:49:43 +08:00
split.cc support Dynamic tensor infer shape and fix memory pool (#176) 2023-11-23 13:11:50 +08:00
squeeze.cc 解除前端对onnx infershape功能的依赖 (#206) 2024-01-12 14:54:27 +08:00
transpose.cc support Dynamic tensor infer shape and fix memory pool (#176) 2023-11-23 13:11:50 +08:00
unary.cc XCCL support (#171) 2024-02-29 11:48:35 +08:00
unsqueeze.cc 解除前端对onnx infershape功能的依赖 (#206) 2024-01-12 14:54:27 +08:00
where.cc support Dynamic tensor infer shape and fix memory pool (#176) 2023-11-23 13:11:50 +08:00