Skip to content
GitLab
Explore
Sign in
Primary navigation
Search or go to…
Project
D
DMML-Replikation
Manage
Activity
Members
Labels
Plan
Issues
Issue boards
Milestones
Wiki
Requirements
Code
Merge requests
Repository
Branches
Commits
Tags
Repository graph
Compare revisions
Snippets
Locked files
Build
Pipelines
Jobs
Pipeline schedules
Test cases
Artifacts
Deploy
Releases
Package registry
Container registry
Model registry
Operate
Environments
Terraform modules
Monitor
Incidents
Service Desk
Analyze
Value stream analytics
Contributor analytics
CI/CD analytics
Repository analytics
Code review analytics
Issue analytics
Insights
Model experiments
Help
Help
Support
GitLab documentation
Compare GitLab plans
Community forum
Contribute to GitLab
Provide feedback
Terms and privacy
Keyboard shortcuts
?
Snippets
Groups
Projects
Show more breadcrumbs
Matthias Keck
DMML-Replikation
Commits
64899f18
Commit
64899f18
authored
3 weeks ago
by
Armin Bacher
Browse files
Options
Downloads
Patches
Plain Diff
Delete run-multi.sh
parent
cf3564c6
Branches
Branches containing commit
No related tags found
No related merge requests found
Changes
1
Hide whitespace changes
Inline
Side-by-side
Showing
1 changed file
Benchmark_Training/run-multi.sh
+0
-28
0 additions, 28 deletions
Benchmark_Training/run-multi.sh
with
0 additions
and
28 deletions
Benchmark_Training/run-multi.sh
deleted
100644 → 0
+
0
−
28
View file @
cf3564c6
#!/bin/bash
#SBATCH --nodes=1
#SBATCH --ntasks=1
#SBATCH --gpus-per-task=2
#SBATCH --gres=gpu:2
#SBATCH --cpus-per-task=16
#SBATCH --mem=64G
#SBATCH --time=02:00:00
# Kopiere das Python-Skript auf den Rechenknoten
sbcast GPT2M_MultiA100.py /zpool1/slurm_data/anhnd/test4.py
# Lokalen Triton-Cache setzen (FlashAttention empfohlen)
export
TRITON_CACHE_DIR
=
/tmp/triton_cache_
$USER
mkdir
-p
"
$TRITON_CACHE_DIR
"
||
export
TRITON_CACHE_DIR
=
$HOME
/.cache/triton
# Set environment for distributed
export
TOKENIZERS_PARALLELISM
=
false
# Optional (für Deepspeed oder fair scale): set NCCL
#export TORCH_DISTRIBUTED_DEBUG=DETAIL
#export NCCL_DEBUG=INFO
#export NCCL_SOCKET_IFNAME=^lo,docker
# Führe das Python-Skript mit absolutem Pfad aus und leite die Ausgabe ins Home-Verzeichnis um
srun
--ntasks
=
1
--gpus-per-task
=
2 torchrun
--nproc_per_node
=
2 /zpool1/slurm_data/anhnd/test4.py flash2
>
~/projekt/GPT2M_2GPU_output_
${
SLURM_JOB_ID
}
.log 2>&1
#Lösche die .py-Datei nach der Berechnung, um Speicherplatz zu sparen
srun
rm
/zpool1/slurm_data/anhnd/test4.py
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