@@ -186,12 +186,12 @@ job_id = $(sbatch --parsable <slurm file>)
186186Job name is used throughout slurm, so change it to something more readable than the script name:
187187
188188```` {.bash }
189- # SBATCH --job-name "AMoreReadableName "
189+ # SBATCH --job-name "A_More_Readable_Name "
190190````
191191
192192<br />
193193
194- ```` {.default filename='#SBATCH --job-name "AMoreReadableName "' code-line-numbers="|7"}
194+ ```` {.default filename='#SBATCH --job-name "A_More_Readable_Name "' code-line-numbers="|7"}
195195[allsoppj@bask-pg0310u18a BasicSlurmFile]$ cat slurm-474832.out
196196
197197This script is running on bask-pg0308u24a.cluster.baskerville.ac.uk
@@ -232,6 +232,33 @@ Don't try <br>
232232
233233Use a template slurm file and substitute values into a new script per job with the ** sed** bash command .
234234
235+
236+ ## Change the hardware you want you job to run on
237+ <br />
238+ Baskerville has two types of GPU,
239+
240+ + ** A100-40** (default)
241+ + ** A100-80**
242+ ```` {.bash}
243+ #SBATCH --constraint=a100_80
244+ ````
245+
246+ ## Change the number of nodes or GPUs
247+ <br />
248+ ```` {.bash}
249+ #SBATCH --gpus 3
250+ #SBATCH --nodes 1
251+ ````
252+ Documented in more detail in the [ docs.baskerville.ac.uk] ( docs.baskerville.ac.uk ) {.external target="_ blank"}
253+
254+ + Leave ** \-\- cpus-per-gpu=36** . Resources such as CPUs and Memory are allocated per GPU.
255+ + All Compute nodes have 4 GPUs.
256+ + So adding more than 4 GPUs on a single node will fail.
257+
258+ See ** 6-MoreResources** to show loading PyTorch and difference between selecting 1 and more GPUs.
259+
260+ + Estimated job start time (worst case) use ** squeue -j=<job ID > --start**
261+
235262## Slurm Arrays - Run many jobs from one Slurm file
236263
237264<br >
@@ -267,29 +294,3 @@ See **2-arrayJobConfig.sh ** for information on loading a different config in ea
267294
268295
269296Used this approach to run nearly 700,000 jobs in blocks of 4000, 500 at a time.
270-
271- ## Change the hardware you want you job to run on
272- <br />
273- Baskerville has two types of GPU,
274-
275- + ** A100-40** (default)
276- + ** A100-80**
277- ```` {.bash}
278- #SBATCH --constraint=a100_80
279- ````
280-
281- ## Change the number of nodes or GPUs
282- <br />
283- ```` {.bash}
284- #SBATCH --gpus 3
285- #SBATCH --nodes 1
286- ````
287- Documented in more detail in the [ docs.baskerville.ac.uk] ( docs.baskerville.ac.uk ) {.external target="_ blank"}
288-
289- + Leave ** \-\- cpus-per-gpu=36** . Resources such as CPUs and Memory are allocated per GPU.
290- + All Compute nodes have 4 GPUs.
291- + So adding more than 4 GPUs on a single node will fail.
292-
293- See ** 6-MoreResources** to show loading PyTorch and difference between selecting 1 and more GPUs.
294-
295- + Estimated job start time (worst case) use ** squeue -j=<job ID > --start**
0 commit comments