-
Notifications
You must be signed in to change notification settings - Fork 0
/
README.setup.biowulf
93 lines (51 loc) · 3.07 KB
/
README.setup.biowulf
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
Author: George Zaki
Email: [email protected]
Date: 04-14-2017
This document includes the instruction to setup the student accounts for the
NIH CANDLE workshop and general instructions to setup the benchmarks on a
general biowulf account.
1. Instructions to setup the student account for the Workshop:
- login to biowulf using your assigned account:
- For example, student1 can login as:
- Run the general setup script:
/data/classes/candle/setup.sh
- Edit the ~/candle-jobscript.sh script to run a specific benchmark
- Submit your job using:
sbatch --partition=student --mem=50g /data/classes/candle/jobrequest.sh
2. Instructions to setup a general biowulf account:
a. Guidelines to submit a batch job:
- Make sure the job request satisfies the following:
- Request at least 50GB of memory for the sbatch job (--mem=50g)
- Request enough time to run the benchmarks as shown in every benchmark's readme.
b. Clone the benchmarks to a directory on Biowulf (make sure you have 30GB of disk space) :
- For example:
export CANDLE_BENCH=/data/`whoami`/candle
git clone https://github.com/ECP-Candle/benchmarks $CANDLE_BENCH
c. Run the benchmark using a singularity container:
- To run the benchmarks, a singularity image that contains all the required
packages is already configured and shared with all biowulf users.
- The image is located at /data/classes/candle/keras.img
I. Use the singularity image in sinteractive job:
- To use the singularity image execute the following commands in your sinteractive job:
sinteractive --mem=50g
module load singularity
- Start the keras.img while binding your $CANDLE_BENCH directory to the /Benchmarks
directory within the singularity image
singularity shell --bind $CANDLE_BENCH:/Benchmarks /data/classes/candle/keras.img
- Run a specific benchmark as shown in /data/classes/candle/candle-jobscript.sh
- For example:
python /Benchmarks/Pilot1/P1B1/p1b1_baseline_keras2.py
II. Use the singularity image in a batch job:
- To execute one or more benchmarks in a batch job, start the singularity
image in an "exec" mode while passing a script of jobs to execute. Check
/data/classes/candle/jobrequest.sh for an example.
- First copy the /data/classes/candle/candle-jobscript.sh to your home directory:
cp /data/classes/candle/candle-jobscript.sh ~/
cp /data/classes/candle/jobrequest.sh ~/candle-jobrequest.sh
- Edit candle-jobscript.sh and uncomment the benchmark(s) you would like to execute.
- Edit ~/candle-jobrequest.sh to bind your $CANDLE_BENCH in your file system to the /Benchmarks
directory in the singularity image. For example, change the singularity call to:
singularity exec --bind $CANDLE_BENCH:/Benchmarks /data/classes/candle/keras.img ./candle-jobscript.sh
- submit the batch job
sbatch --mem=50g ~/candle-jobrequest.sh