Skip to content

Commit 690233c

Browse files
sharathtsnv-kkudrynski
authored andcommitted
[BERT/PyT] Config update
1 parent 2132ae7 commit 690233c

File tree

9 files changed

+131
-684
lines changed

9 files changed

+131
-684
lines changed

PyTorch/LanguageModeling/BERT/README.md

+5-4
Original file line numberDiff line numberDiff line change
@@ -344,10 +344,10 @@ out_dir=${5:-"/workspace/bert/results/SST-2"} # For SST-2.
344344
This repository contains a number of predefined configurations to run the SQuAD, GLUE and pre-training on NVIDIA DGX-1, NVIDIA DGX-2H or NVIDIA DGX A100 nodes in `scripts/configs/squad_config.sh`, `scripts/configs/glue_config.sh` and `scripts/configs/pretrain_config.sh`. For example, to use the default DGX A100 8 gpu config, run:
345345

346346
```
347-
bash scripts/run_squad.sh $(source scripts/configs/squad_config.sh && dgxa10080g_8gpu_fp16_p2binned) # For the SQuAD v1.1 dataset.
348-
bash scripts/run_glue.sh $(source scripts/configs/glue_config.sh && mrpc_dgxa10080g_8gpu_fp16_p2binned) # For the MRPC dataset.
349-
bash scripts/run_glue.sh $(source scripts/configs/glue_config.sh && sst-2_dgxa10080g_8gpu_fp16_p2binned) # For the SST-2 dataset.
350-
bash scripts/run_pretraining.sh $(source scripts/configs/pretrain_config.sh && dgxa10080g_8gpu_fp16) # For pre-training
347+
bash scripts/run_squad.sh $(source scripts/configs/squad_config.sh && dgxa100-80g_8gpu_fp16) # For the SQuAD v1.1 dataset.
348+
bash scripts/run_glue.sh $(source scripts/configs/glue_config.sh && mrpc_dgxa100-80g_8gpu_fp16) # For the MRPC dataset.
349+
bash scripts/run_glue.sh $(source scripts/configs/glue_config.sh && sst-2_dgxa100-80g_8gpu_fp16) # For the SST-2 dataset.
350+
bash scripts/run_pretraining.sh $(source scripts/configs/pretrain_config.sh && dgxa100-80g_8gpu_fp16) # For pre-training
351351
```
352352

353353
## Advanced
@@ -891,6 +891,7 @@ January 2022
891891
- Knowledge Distillation support
892892
- Pre-training with native AMP, native DDP, and TorchScript with NVFuser backend
893893
- Pre-training using [Language Datasets and Data Loaders (LDDL)](../../../Tools/lddl)
894+
- Binned pretraining for phase2 with LDDL using a bin size of 64
894895

895896
July 2020
896897
- Updated accuracy and performance tables to include A100 results
Loading

PyTorch/LanguageModeling/BERT/lddl/setup.py

+10-10
Original file line numberDiff line numberDiff line change
@@ -22,16 +22,16 @@
2222
packages=find_packages(),
2323
python_requires='>=3.6',
2424
install_requires=[
25-
'dask[complete]>=2021.2.0',
26-
'distributed>=2021.2.0',
27-
'dask-mpi>=2.21.0',
28-
'pyarrow>=3.0.0',
29-
'mpi4py>=3.0.3',
30-
'transformers>=4.3.2',
31-
'wikiextractor>=3.0.5',
32-
'news-please>=1.5.18',
33-
'cchardet>=2.1.7',
34-
'awscli>=1.19.53',
25+
'dask[complete]==2021.7.1',
26+
'distributed==2021.7.1',
27+
'dask-mpi==2021.11.0',
28+
'pyarrow==4.0.1',
29+
'mpi4py==3.1.3',
30+
'transformers==4.16.2',
31+
'wikiextractor==3.0.6',
32+
'news-please==1.5.21',
33+
'cchardet==2.1.7',
34+
'awscli==1.22.55',
3535
'wikiextractor @ git+https://github.com/attardi/wikiextractor.git',
3636
],
3737
entry_points={

0 commit comments

Comments
 (0)