lucasdelimanogueira
|
70cbd45ae7
|
fix memory usage
|
2024-06-05 21:04:09 -03:00 |
|
lucasdelimanogueira
|
e6d58121f1
|
fix memory usage
|
2024-06-05 20:24:09 -03:00 |
|
lucasdelimanogueira
|
e670559fe4
|
fix memory issues del tensors
|
2024-06-05 15:43:19 -03:00 |
|
lucasdelimanogueira
|
6701179c22
|
del tensor free memory
|
2024-06-05 14:31:48 -03:00 |
|
lucasdelimanogueira
|
b56ed81cd7
|
add comment on reduce cuda
|
2024-06-04 11:21:56 -03:00 |
|
lucasdelimanogueira
|
64f08a8963
|
multigpu training support v0.0.5
|
2024-06-03 15:37:57 -03:00 |
|
lucasdelimanogueira
|
e706b81007
|
fix broadcast and reduce gradients distributeddataparallel and tensor hooks
|
2024-06-03 11:40:05 -03:00 |
|
lucasdelimanogueira
|
4b683a20e9
|
fix distributed ncclCommInitRank
|
2024-05-31 19:40:58 -03:00 |
|
lucasdelimanogueira
|
38a860eba1
|
fix send to other devices and compile on cloud
|
2024-05-29 17:48:37 -03:00 |
|
lucasdelimanogueira
|
002287d832
|
remove print from cuda to cpu
|
2024-05-27 20:55:50 -03:00 |
|
lucasdelimanogueira
|
0e9c4779a8
|
add argument src to broadcast distributed
|
2024-05-25 09:38:14 -03:00 |
|
lucasdelimanogueira
|
1e0c408ffb
|
fix import distributed nn
|
2024-05-25 08:57:35 -03:00 |
|
lucasdelimanogueira
|
5283c1bc06
|
allreduce mean
|
2024-05-25 08:52:20 -03:00 |
|
lucasdelimanogueira
|
29ee569d04
|
fix import distributed
|
2024-05-24 19:55:19 -03:00 |
|
lucasdelimanogueira
|
94b3582a46
|
test broadcast and all reduce
|
2024-05-24 18:31:59 -03:00 |
|
lucasdelimanogueira
|
55db13836b
|
broadcast allreduce nccl
|
2024-05-24 18:19:39 -03:00 |
|
lucasdelimanogueira
|
e55c1195ec
|
implemented mpi for distributed run
|
2024-05-24 15:12:53 -03:00 |
|
lucasdelimanogueira
|
b152ae2a1d
|
first commit distributed module
|
2024-05-23 19:15:30 -03:00 |
|
lucasdelimanogueira
|
a0bfd6a0eb
|
small fix 3D transpose
|
2024-05-23 02:34:38 -03:00 |
|
lucasdelimanogueira
|
38e629fc38
|
make contiguous cuda
|
2024-05-23 02:27:46 -03:00 |
|
lucasdelimanogueira
|
702ebaf87c
|
1D, 2D and 3D transpose cuda
|
2024-05-23 02:09:32 -03:00 |
|
lucasdelimanogueira
|
fb6429aad6
|
min cuda and min cuda axis
|
2024-05-23 01:39:37 -03:00 |
|
lucasdelimanogueira
|
223229fcd3
|
min cuda and min cuda axis
|
2024-05-23 01:35:45 -03:00 |
|
lucasdelimanogueira
|
e2a0262c4e
|
small fix max cuda axis was setting init to 0 instead of -inf
|
2024-05-23 01:29:23 -03:00 |
|
lucasdelimanogueira
|
5882f1d801
|
max cuda and max cuda axis
|
2024-05-23 01:26:28 -03:00 |
|
lucasdelimanogueira
|
f25a060294
|
fix sum cuda axis with atomicAdd
|
2024-05-23 01:06:21 -03:00 |
|
lucasdelimanogueira
|
c02170731f
|
remove atomicAdd test
|
2024-05-23 00:38:33 -03:00 |
|
lucasdelimanogueira
|
192a50f773
|
fix sum cuda axis atomicAdd
|
2024-05-23 00:30:48 -03:00 |
|
lucasdelimanogueira
|
b8ad9ef282
|
fix sum cuda memory corruption
|
2024-05-23 00:03:26 -03:00 |
|
lucasdelimanogueira
|
d093015d15
|
fix sum cuda axis
|
2024-05-22 21:07:09 -03:00 |
|
lucasdelimanogueira
|
2c7571c371
|
small fix sum cuda
|
2024-05-22 20:59:46 -03:00 |
|
lucasdelimanogueira
|
950cde13f7
|
small fix sum cuda
|
2024-05-22 20:59:25 -03:00 |
|
lucasdelimanogueira
|
718a4c9823
|
sum axis cuda version 2
|
2024-05-22 19:55:00 -03:00 |
|
lucasdelimanogueira
|
f446daf707
|
sum axis cuda version
|
2024-05-22 19:32:47 -03:00 |
|
lucasdelimanogueira
|
65dc86aaa4
|
broadcasted batched matmul cuda
|
2024-05-22 19:20:54 -03:00 |
|
lucasdelimanogueira
|
cebf890165
|
sigmoid kernel avoid overflow
|
2024-05-22 13:24:01 -03:00 |
|
lucasdelimanogueira
|
a39bd59a3f
|
softmax operation
|
2024-05-21 01:22:38 -03:00 |
|
lucasdelimanogueira
|
fd831f950d
|
fix add min max keepdim
|
2024-05-21 00:38:34 -03:00 |
|
lucasdelimanogueira
|
5e72fe57d3
|
softmax v1 need fix some erros yet
|
2024-05-20 20:16:41 -03:00 |
|
lucasdelimanogueira
|
b324b7e1fa
|
broadcasted equal operation
|
2024-05-20 16:42:45 -03:00 |
|
lucasdelimanogueira
|
cc483417dc
|
max min and equal operations
|
2024-05-20 12:50:45 -03:00 |
|
lucasdelimanogueira
|
2df6c36d99
|
fixed sum axis with keepdim
|
2024-05-17 19:29:16 -03:00 |
|
lucasdelimanogueira
|
5dd3f3d202
|
implementing keepdim sum
|
2024-05-17 19:08:09 -03:00 |
|
lucasdelimanogueira
|
de72cd67ba
|
sub broadcasted cuda
|
2024-05-17 13:15:33 -03:00 |
|
lucasdelimanogueira
|
64fb3f9edb
|
fix sum axis autograd and reversed add broadcasted order
|
2024-05-17 11:39:26 -03:00 |
|
lucasdelimanogueira
|
b3c0543114
|
batched matmul cuda
|
2024-05-16 19:53:39 -03:00 |
|
lucasdelimanogueira
|
5da5a7ad0d
|
batched matmul cuda
|
2024-05-16 19:34:58 -03:00 |
|
lucasdelimanogueira
|
c11cfd7f22
|
fix transpose cuda
|
2024-05-16 19:10:22 -03:00 |
|
lucasdelimanogueira
|
85d8e74ad1
|
fix cuda add broadcasted
|
2024-05-16 18:06:50 -03:00 |
|
lucasdelimanogueira
|
54a5f81b3e
|
add broadcasted cuda
|
2024-05-16 17:23:38 -03:00 |
|