Skip to content

Commit f0d6c73

Browse files
committed
Merge branch 'develop' into issue-119
2 parents ee09079 + a9bf3f3 commit f0d6c73

11 files changed

+7627
-51
lines changed

Pilot1/ST1/VocabFiles_spe/SPE_ChEMBL.txt

Lines changed: 3002 additions & 0 deletions
Large diffs are not rendered by default.

Pilot1/ST1/VocabFiles_spe/vocab_spe.txt

Lines changed: 3132 additions & 0 deletions
Large diffs are not rendered by default.
Lines changed: 52 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,52 @@
1+
{
2+
"general": {
3+
"use_hvd": true,
4+
"batch_size": 64,
5+
"epochs": 400,
6+
"lr": 0.00000991301767144166,
7+
"loss_fn": "mean_squared_error"
8+
},
9+
10+
"data_loading": {
11+
"data_path": "/lus/grand/projects/datascience/avasan/Data_Docking/2M-flatten",
12+
"rec": "3CLPro_7BQY_A_1_F",
13+
"pattern": "Orderable_zinc_db_enaHLL.sorted.4col.descriptors.parquet.xform-smiles.csv.reg"
14+
},
15+
16+
"tokenization": {
17+
"vocab_size": 3132,
18+
"maxlen": 45,
19+
"tokenizer": {
20+
"category": "smilespair",
21+
"spe_file": "VocabFiles/SPE_ChEMBL.txt",
22+
"vocab_file": "VocabFiles/vocab_spe.txt"
23+
}
24+
},
25+
26+
"architecture": {
27+
"embedding": {
28+
"embed_dim": 128
29+
},
30+
"transformer_block": {
31+
"num_blocks": 5,
32+
"activation": "selu",
33+
"ff_dim": 128,
34+
"num_heads": 21,
35+
"dr1": 0.12717945391278226,
36+
"dr2": 0.12717945391278226,
37+
"drop_mha": true
38+
},
39+
"regressor_head": {
40+
"activation": "selu",
41+
"dr": 0.04990303516069576
42+
}
43+
},
44+
45+
"callbacks": {
46+
"checkpt_file": "smile_regress.autosave.model.h5",
47+
"log_csv": "smile_regress.training.log",
48+
"patience_red_lr": 20,
49+
"patience_early_stop": 100
50+
}
51+
52+
}
Lines changed: 27 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,27 @@
1+
#!/bin/bash
2+
#PBS -N st_spe
3+
#PBS -l select=4
4+
#PBS -l walltime=12:00:00
5+
#PBS -q preemptable
6+
#PBS -l filesystems=grand
7+
#PBS -A datascience
8+
#PBS -o logs/
9+
#PBS -e logs/
10+
#PBS -m abe
11+
12+
13+
module load conda/2022-09-08
14+
conda activate
15+
16+
cd /grand/datascience/avasan/ST_Benchmarks/Test_Tokenizers/SMILESPair_Encoder_continue
17+
18+
NP=16
19+
PPN=4
20+
OUT=logfile.log
21+
let NDEPTH=64/$NP
22+
let NTHREADS=$NDEPTH
23+
24+
TF_GPU_ALLOCATOR=cuda_malloc_async
25+
export TF_FORCE_GPU_ALLOW_GROWTH=true
26+
27+
mpiexec --np 16 -ppn 4 --cpu-bind verbose,list:0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16 -env NCCL_COLLNET_ENABLE=1 -env NCCL_NET_GDR_LEVEL=PHB python smiles_regress_transformer_run.py > $OUT

0 commit comments

Comments
 (0)