Nigel
/
mindspore
forked from mindspore-Ecosystem/mindspore
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Add ascend memory for pangu model

This commit is contained in:
huangxinjing 2021-07-12 11:35:27 +08:00
parent 5491b2c982
commit 3211e527ad
2 changed files with 3 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
model_zoo/official/nlp/pangu_alpha/README.md
View File

@ -120,7 +120,7 @@ The vocab size of `vocab.vocab` is 40000, and the `eod id` is 6.
### Training On Ascend
Currently the scripts provide three default configures : `2.6B` `13B` and `200B`. The following command will start training `2.6B` model on 8 **Ascend cards**.
Currently the scripts provide three default configures : `2.6B` `13B` and `200B`.
```bash
@ -465,7 +465,7 @@ Please check the official [homepage](https://gitee.com/mindspore/mindspore/tree/
# [Requirements](#contents)
- mindspore 1.2.1 or higher version
- mindspore 1.3.0
- jieba 0.42.1
- sentencepiece 0.1.94
- transformers >= 4.7.0

2
model_zoo/official/nlp/pangu_alpha/train.py
View File

@ -83,7 +83,7 @@ def run_train(args_opt):
# Set execution mode
context.set_context(mode=context.GRAPH_MODE,
device_target=args_opt.device_target)
context.set_context(variable_memory_max_size="30GB")
context.set_context(variable_memory_max_size="31GB")
# Set parallel context
if args_opt.distribute == "true":
D.init()