issues
search
ROCm
/
apex
A PyTorch Extension: Tools for easy mixed precision and distributed training in Pytorch
BSD 3-Clause "New" or "Revised" License
17
stars
14
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
support megatron seq_len > 4096
#135
ramcherukuri
opened
4 days ago
0
test_mlp benchmark got accuracy assert error
#134
ZJLi2013
opened
2 weeks ago
0
change compute type for F16 wrapper around cublas GEMMEx
#133
suachong
closed
1 month ago
1
Errors when building on an MI250x server with ROCm 5.7 and PyTorch 2.2.1
#132
netw0rkf10w
opened
4 months ago
3
Moving master to version 1.3.0
#131
ramcherukuri
closed
5 months ago
0
Add setting of env flag when apex is turned on
#130
pragupta
closed
5 months ago
0
Batchnorm support
#129
ramcherukuri
closed
5 months ago
0
CI Test do not merge
#128
JnProfile
closed
6 months ago
0
moving from rocBLAS to hipBLAS
#127
ramcherukuri
closed
5 months ago
3
Moving version to 1.2.0
#126
pruthvistony
closed
7 months ago
0
CP of hipblas, README.md related changes
#125
pruthvistony
closed
7 months ago
2
Rel1.1.0 cherrypick master
#124
ramcherukuri
closed
7 months ago
3
Rel1.1.0 cherrypick master
#123
ramcherukuri
closed
7 months ago
0
remove HCC references
#122
jeffdaily
closed
8 months ago
2
Adding version.txt with 1.1.0
#121
ramcherukuri
closed
8 months ago
0
Adding version.txt with 1.1.0 to master
#120
ramcherukuri
closed
9 months ago
0
syncing with Nvidia-master
#119
ramcherukuri
closed
9 months ago
0
Is RoCm apex.amp deprecated & behavior mismatch vs NVIDIA APEX
#118
fxmarty
opened
10 months ago
1
Update version from 0.1
#117
fxmarty
opened
10 months ago
0
Revert "Changes to support hipblas migration (#113)"
#116
pruthvistony
closed
10 months ago
0
Problems building apex with ROCm-5.4, 5.5, and 5.6
#115
adammoody
opened
10 months ago
3
Add setting of env flag when apex is turned on
#114
alugorey
opened
11 months ago
2
Changes to support hipblas migration
#113
pruthvistony
closed
11 months ago
0
Adding pyproject.toml file
#112
pruthvistony
closed
1 year ago
1
Updated pip no longer supports `--install-option` for building without cloning
#111
loadams
closed
1 year ago
1
Update rccl header include path
#110
pruthvistony
closed
1 year ago
0
Add FusedLARS optimizer
#109
luise1030
closed
1 year ago
0
Add FusedLARS optimizer
#108
hubertlu-tw
closed
1 year ago
0
Cherry-picks some commits to replace torch.Tensor and remove dependency on six
#107
hubertlu-tw
closed
1 year ago
1
no module torch._six
#106
zstreet87
opened
1 year ago
1
Luise/gbn optimization
#105
luise1030
closed
1 year ago
0
Grid optimization - Chunk_Size optimization.
#104
aspanday
closed
1 year ago
3
Updating BLOCK_SIZE to 1024 in all optimizers.
#103
aspanday
closed
1 year ago
0
Fatal error: 'cuda_runtime_api.h' file not found
#102
lvcc2018
opened
1 year ago
5
Support all the softmax extensions and cherry-pick transformer-related commits
#101
hubertlu-tw
opened
1 year ago
0
Update register keyword handling for C++17
#100
pruthvistony
closed
1 year ago
0
Fix a bug in fused_dense_cuda on ROCm
#99
hubertlu-tw
closed
1 year ago
0
Unskip some unit tests related to issue #82
#98
hubertlu-tw
closed
1 year ago
1
Consider both contiguous and channels_last tensors for FusedSGD
#97
hubertlu-tw
closed
1 year ago
1
Make index_mul_2d extension backward compatible for Atomic header include
#96
hubertlu-tw
closed
1 year ago
1
Faster build using ninja
#95
hubertlu-tw
closed
1 year ago
2
Enable --fast_layer_norm for ROCm
#94
hubertlu-tw
opened
1 year ago
3
Cherry-pick fused_adam_cuda related patches and add its unit tests to the ROCm extension test script
#93
hubertlu-tw
opened
1 year ago
3
Failing tests in --peer_memory
#92
hubertlu-tw
opened
1 year ago
0
Enable --focal_loss and --index_mul_2d extensions for ROCm
#91
hubertlu-tw
closed
1 year ago
1
cached cast fix
#90
hubertlu-tw
closed
1 year ago
3
The failing unit tests in test_transducer_joint.py
#89
hubertlu-tw
opened
1 year ago
0
Enable --transducer extension for ROCm
#88
hubertlu-tw
closed
1 year ago
5
Enable --peer_memory and --nccl p2p extensions for ROCm
#87
hubertlu-tw
closed
1 year ago
2
Add a wrapper to skip flaky tests and un-skip some MLP unit tests
#86
hubertlu-tw
closed
1 year ago
1
Next