Skip to content
GitLab
Explore
Sign in
Primary navigation
Search or go to…
Project
2
2023-nov-intro-to-supercompting-jsc
Manage
Activity
Members
Labels
Plan
Issues
Issue boards
Milestones
Wiki
Code
Merge requests
Repository
Branches
Commits
Tags
Repository graph
Compare revisions
Snippets
Build
Pipelines
Jobs
Pipeline schedules
Artifacts
Deploy
Releases
Package registry
Container registry
Model registry
Operate
Environments
Terraform modules
Monitor
Incidents
Analyze
Value stream analytics
Contributor analytics
CI/CD analytics
Repository analytics
Model experiments
Help
Help
Support
GitLab documentation
Compare GitLab plans
GitLab community forum
Contribute to GitLab
Provide feedback
Keyboard shortcuts
?
Snippets
Groups
Projects
Show more breadcrumbs
Alexandre Strube
2023-nov-intro-to-supercompting-jsc
Commits
88132c8c
Commit
88132c8c
authored
May 30, 2023
by
Alexandre Strube
Browse files
Options
Downloads
Patches
Plain Diff
fix git stuff
parent
2a23d51f
Branches
master
Branches containing commit
No related tags found
No related merge requests found
Pipeline
#140724
passed
May 30, 2023
Stage: test
Stage: deploy
Changes
3
Pipelines
1
Show whitespace changes
Inline
Side-by-side
Showing
3 changed files
01-deep-learning-on-supercomputers.md
+0
-2
0 additions, 2 deletions
01-deep-learning-on-supercomputers.md
public/01-deep-learning-on-supercomputers.html
+0
-3
0 additions, 3 deletions
public/01-deep-learning-on-supercomputers.html
src/distrib.slurm
+5
-11
5 additions, 11 deletions
src/distrib.slurm
with
5 additions
and
16 deletions
01-deep-learning-on-supercomputers.md
+
0
−
2
View file @
88132c8c
...
@@ -378,8 +378,6 @@ date: May 31, 2023
...
@@ -378,8 +378,6 @@ date: May 31, 2023
-
Let's take a simple model
-
Let's take a simple model
-
Run it "serially" (single-gpu)
-
Run it "serially" (single-gpu)
-
Then make it data parallel among multiple gpus in one node
-
Then make it data parallel among multiple gpus in one node
-
Afterwards, make it data parallel among multiple nodes
-
Finally, make it in model parallel
---
---
...
...
This diff is collapsed.
Click to expand it.
public/01-deep-learning-on-supercomputers.html
+
0
−
3
View file @
88132c8c
...
@@ -591,9 +591,6 @@ gpus</li>
...
@@ -591,9 +591,6 @@ gpus</li>
<li
class=
"fragment"
>
Run it “serially” (single-gpu)
</li>
<li
class=
"fragment"
>
Run it “serially” (single-gpu)
</li>
<li
class=
"fragment"
>
Then make it data parallel among multiple gpus in
<li
class=
"fragment"
>
Then make it data parallel among multiple gpus in
one node
</li>
one node
</li>
<li
class=
"fragment"
>
Afterwards, make it data parallel among multiple
nodes
</li>
<li
class=
"fragment"
>
Finally, make it in model parallel
</li>
</ul>
</ul>
</section>
</section>
<section
id=
"expected-imports"
class=
"slide level2"
>
<section
id=
"expected-imports"
class=
"slide level2"
>
...
...
This diff is collapsed.
Click to expand it.
src/distrib.slurm
+
5
−
11
View file @
88132c8c
#!/bin/bash -x
#!/bin/bash -x
#SBATCH --account=training2306
#SBATCH --account=training2306
#SBATCH --nodes=
1
#SBATCH --nodes=
2
#SBATCH --job-name=ai-multi-gpu
#SBATCH --job-name=ai-multi-gpu
#SBATCH --ntasks-per-node=1
#SBATCH --ntasks-per-node=1
#SBATCH --cpus-per-task=48
#SBATCH --cpus-per-task=48
...
@@ -18,26 +18,21 @@ MASTER_ADDR="$(scontrol show hostnames "$SLURM_JOB_NODELIST" | head -n 1)"
...
@@ -18,26 +18,21 @@ MASTER_ADDR="$(scontrol show hostnames "$SLURM_JOB_NODELIST" | head -n 1)"
MASTER_ADDR
=
"
${
MASTER_ADDR
}
i"
MASTER_ADDR
=
"
${
MASTER_ADDR
}
i"
# Get IP for hostname.
# Get IP for hostname.
export
MASTER_ADDR
=
"
$(
nslookup
"
$MASTER_ADDR
"
|
grep
-oP
'(?<=Address: ).*'
)
"
export
MASTER_ADDR
=
"
$(
nslookup
"
$MASTER_ADDR
"
|
grep
-oP
'(?<=Address: ).*'
)
"
<<<<<<
< HEAD
export
MASTER_PORT
=
7010
export
MASTER_PORT
=
7010
export
GPUS_PER_NODE
=
4
export
GPUS_PER_NODE
=
4
export
NNODES
=
$SLURM_JOB_NUM_NODES
export
NNODES
=
$SLURM_JOB_NUM_NODES
# do not remove or the training will hang and nodes will be lost w/o this workaround
# do not remove or the training will hang and nodes will be lost w/o this workaround
#
export CUDA_LAUNCH_BLOCKING=1
export
CUDA_LAUNCH_BLOCKING
=
1
# hide duplicated errors using this hack - will be properly fixed in pt-1.12
# hide duplicated errors using this hack - will be properly fixed in pt-1.12
#
export TORCHELASTIC_ERROR_FILE=/tmp/torch-elastic-error.json
export
TORCHELASTIC_ERROR_FILE
=
/tmp/torch-elastic-error.json
# force crashing on nccl issues like hanging broadcast
# force crashing on nccl issues like hanging broadcast
#
export NCCL_ASYNC_ERROR_HANDLING=1
export
NCCL_ASYNC_ERROR_HANDLING
=
1
# handle timeouts
# handle timeouts
export
NCCL_IB_TIMEOUT
=
20
export
NCCL_IB_TIMEOUT
=
20
=======
export
MASTER_PORT
=
6000
GPUS_PER_NODE
=
4
NNODES
=
$SLURM_JOB_NUM_NODES
>>>>>>>
423694b75394e474f9d309661bdceafb44c402e0
# Make sure we are on the right directory
# Make sure we are on the right directory
cd
$HOME
/2023-may-intro-to-supercompting-jsc/src
cd
$HOME
/2023-may-intro-to-supercompting-jsc/src
...
@@ -45,7 +40,6 @@ cd $HOME/2023-may-intro-to-supercompting-jsc/src
...
@@ -45,7 +40,6 @@ cd $HOME/2023-may-intro-to-supercompting-jsc/src
# This loads modules and python packages
# This loads modules and python packages
source
sc_venv_template/activate.sh
source
sc_venv_template/activate.sh
export
LOGLEVEL
=
INFO
export
LOGLEVEL
=
INFO
# Run the demo
# Run the demo
time
srun bash
-c
'accelerate launch \
time
srun bash
-c
'accelerate launch \
...
...
This diff is collapsed.
Click to expand it.
Preview
0%
Loading
Try again
or
attach a new file
.
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Save comment
Cancel
Please
register
or
sign in
to comment