Difference between revisions of "Eosc/easybuild/benchmark"

From Nordic Language Processing Laboratory
Jump to: navigation, search
(TensorFlow)
(TensorFlow)
Line 32: Line 32:
 
We use training a toy BERT model on 4 GPUs as a benchmark. After installing the NLPL software stack, load the '''NLPL-nvidia_BERT''' module (see below) and run the following command:
 
We use training a toy BERT model on 4 GPUs as a benchmark. After installing the NLPL software stack, load the '''NLPL-nvidia_BERT''' module (see below) and run the following command:
  
  <nowiki>$ train_bert.sh CORPUS VOCAB CONFIG </nowiki>,
+
  <nowiki>$ train_bert.sh CORPUS VOCAB CONFIG </nowiki>
  
 
where CORPUS is a path to a directory with text files, VOCAB is a path to a WordPiece vocabulary, CONFIG is a path to a BERT configuration JSON (defining the model hyperparameters).
 
where CORPUS is a path to a directory with text files, VOCAB is a path to a WordPiece vocabulary, CONFIG is a path to a BERT configuration JSON (defining the model hyperparameters).

Revision as of 02:46, 27 November 2020

Background

In the context of the EOSC-Nordic EasyBuild pilot, the following page provides instructions for how to benchmark different software configurations on typical problems that are likely to affect NLPL users. Relevant variation will typically contrast pre-compiled binary installations (e.g. `pip` wheels) vs. locally compiled modules, where architecture-specific optimizations are enabled and optimized libraries (e.g. Intel MKL) are used.

NumPy

We use this Python script which runs multiple random matrix multiplications and singular value decompositions (SVD). Only CPU is employed.

OpenBLAS on Saga

$ module use -a /cluster/shared/nlpl/software/easybuild_ak/easybuild/install/modules/all/
$ module load NLPL-numpy/1.18.1-foss-2019b-Python-3.7.4
$ python3 tests/numpy/numpy_test.py
Multiplication took 78 seconds.
SVD took 60 seconds.

IMKL on Saga

$ module use -a /cluster/shared/nlpl/software/easybuild_ak/easybuild/install/modules/all/
$ module load NLPL-numpy/1.18.1-gomkl-2019b-Python-3.7.4
$ python3 tests/numpy/numpy_test.py
Multiplication took 55 seconds.
SVD took 49 seconds.

TensorFlow

We use training a toy BERT model on 4 GPUs as a benchmark. After installing the NLPL software stack, load the NLPL-nvidia_BERT module (see below) and run the following command:

$ train_bert.sh CORPUS VOCAB CONFIG 

where CORPUS is a path to a directory with text files, VOCAB is a path to a WordPiece vocabulary, CONFIG is a path to a BERT configuration JSON (defining the model hyperparameters).

Ready-to-use toy data for Norwegian can be downloaded here, but in principle any plain text corpus can be fed to this code.

TF with OpenBLAS on Saga

module load NLPL-nvidia_BERT/20.06.8-foss-2019b-TensorFlow-1.15.2-Python-3.7.4
$ train_bert.sh no_wiki/ norwegian_wordpiece_vocab_20k.txt norbert_config.json

Training time: 00:46:27

TF with OpenMKL on Saga

module load NLPL-nvidia_BERT/20.06.8-gomkl-2019b-TensorFlow-1.15.2-Python-3.7.4
$ train_bert.sh no_wiki/ norwegian_wordpiece_vocab_20k.txt norbert_config.json

Training time: 00:46:19