Add a Dockerfile for PyTorch + ROCm based on official AMD released ar…
48d3efbd
Add a new artifact single-amdgpu testing on main
c1acac08
Attempt to test the workflow without merging.
70dbee0f
Changed BERT to check if things are triggered
96639bb5
Meet the dependencies graph on workflow
8f3e698f
Revert BERT changes
4cd38711
Add check_runners_amdgpu to correctly mount and check availability
cc62d3dd
Rename setup to setup_gpu for CUDA and add setup_amdgpu for AMD
a4a639c7
Fix all the needs.setup -> needs.setup_[gpu|amdgpu] dependencies
24853839
Fix setup dependency graph to use check_runner_amdgpu
f99374d4
Let's do the runner status check only on AMDGPU target
c045c1e4
Update the Dockerfile.amd to put ourselves in / rather than /var/lib
d3c3e728
Restore the whole setup for CUDA too.
d2372270
Let's redisable them
1a0e3020
Change BERT to trigger tests
ad48aeb1
Restore BERT
0107f558
Add torchaudio with rocm 5.6 to AMD Dockerfile (#26050)
4a2efa4e
Place AMD GPU tests in a separate workflow (correct branch) (#26105)
cd106b49
Fix invalid job name is dependencies.
933b00f1
Remove tests multi-amdgpu for now.
7c1edd9e
Use single-amdgpu
4c359796
Use --net=host for now.
8dcc3b4b
Remote host networking.
17e07f58
Removed duplicated check_runners_amdgpu step
d76455d8
Let's tag machine-types with mi210 for now.
f58b7aef
Machine type should be only mi210
422110e4
Remove unnecessary push.branches item
6b860bed
Apply review suggestions moving from `x-amdgpu` to `x-gpu` introducin…
a8690cdb
Remove amdgpu from step names.
ec4787f1
finalize
047dd96d
ydshieh
changed the title
ydshieh
approved these changes
on 2023-09-20
delete
d1fb120f
LysandreJik
approved these changes
on 2023-09-20
Login to write a write a comment.
Login via GitHub