p83571920
/
mindspore
forked from mindspore-Ecosystem/mindspore
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

!11086 fix fasttext doc issue 

From: @zhaojichen
Reviewed-by: @wuxuejian,@liangchenghui
Signed-off-by: @liangchenghui
This commit is contained in:
mindspore-ci-bot 2021-01-09 10:09:05 +08:00 committed by Gitee
parent a76e2b2649 756f67151b
commit ee1c80abc6
5 changed files with 29 additions and 27 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

45
model_zoo/official/nlp/fasttext/README.md
View File

@ -64,13 +64,13 @@ After dataset preparation, you can start training and evaluation as follows:
```bash ```bash
# run training example # run training example
cd ./scripts cd ./scripts
sh run_standalone_train.sh [TRAIN_DATASET] sh run_standalone_train.sh [TRAIN_DATASET] [DEVICEID]
# run distributed training example # run distributed training example
sh run_distribute_train.sh [TRAIN_DATASET] [RANK_TABLE_PATH] sh run_distribute_train.sh [TRAIN_DATASET] [RANK_TABLE_PATH]
# run evaluation example # run evaluation example
sh run_eval.sh [EVAL_DATASET_PATH] [DATASET_NAME] [MODEL_CKPT] sh run_eval.sh [EVAL_DATASET_PATH] [DATASET_NAME] [MODEL_CKPT] [DEVICEID]
``` ```
# [Script Description](#content) # [Script Description](#content)
@ -116,6 +116,7 @@ Parameters for both training and evaluation can be set in config.py. All the dat
```text ```text
vocab_size # vocabulary size. vocab_size # vocabulary size.
buckets # bucket sequence length. buckets # bucket sequence length.
test_buckets # test dataset bucket sequence length
batch_size # batch size of input dataset. batch_size # batch size of input dataset.
embedding_dims # The size of each embedding vector. embedding_dims # The size of each embedding vector.
num_class # number of labels. num_class # number of labels.
@ -134,7 +135,7 @@ Parameters for both training and evaluation can be set in config.py. All the dat
```bash ```bash
cd ./scripts cd ./scripts
sh run_standalone_train.sh [DATASET_PATH] sh run_standalone_train.sh [DATASET_PATH] [DEVICEID]
``` ```
- Running scripts for distributed training of FastText. Task training on multiple device and run the following command in bash to be executed in `scripts/`: - Running scripts for distributed training of FastText. Task training on multiple device and run the following command in bash to be executed in `scripts/`:
@ -150,7 +151,7 @@ Parameters for both training and evaluation can be set in config.py. All the dat
``` bash ``` bash
cd ./scripts cd ./scripts
sh run_eval.sh [DATASET_PATH] [DATASET_NAME] [MODEL_CKPT] sh run_eval.sh [DATASET_PATH] [DATASET_NAME] [MODEL_CKPT] [DEVICEID]
``` ```
Note: The `DATASET_PATH` is path to mindrecord. eg. /dataset_path/*.mindrecord Note: The `DATASET_PATH` is path to mindrecord. eg. /dataset_path/*.mindrecord
@ -167,13 +168,13 @@ Parameters for both training and evaluation can be set in config.py. All the dat
| uploaded Date | 12/21/2020 (month/day/year) | | uploaded Date | 12/21/2020 (month/day/year) |
| MindSpore Version | 1.1.0 | | MindSpore Version | 1.1.0 |
| Dataset | AG's News Topic Classification Dataset | | Dataset | AG's News Topic Classification Dataset |
| Training Parameters | epoch=5, batch_size=128 | | Training Parameters | epoch=5, batch_size=512 |
| Optimizer | Adam | | Optimizer | Adam |
| Loss Function | Softmax Cross Entropy | | Loss Function | Softmax Cross Entropy |
| outputs | probability | | outputs | probability |
| Speed | 112ms/step (8pcs) | | Speed | 10ms/step (1pcs) |
| Total Time | 66s (8pcs) | | Epoch Time | 2.36s (1pcs) |
| Loss | 0.00082 | | Loss | 0.0067 |
| Params (M) | 22 | | Params (M) | 22 |
| Checkpoint for inference | 254M (.ckpt file) | | Checkpoint for inference | 254M (.ckpt file) |
| Scripts | [fasttext](https://gitee.com/mindspore/mindspore/tree/master/model_zoo/official/nlp/fasttext) | | Scripts | [fasttext](https://gitee.com/mindspore/mindspore/tree/master/model_zoo/official/nlp/fasttext) |
@ -184,13 +185,13 @@ Parameters for both training and evaluation can be set in config.py. All the dat
| uploaded Date | 11/21/2020 (month/day/year) | | uploaded Date | 11/21/2020 (month/day/year) |
| MindSpore Version | 1.1.0 | | MindSpore Version | 1.1.0 |
| Dataset | DBPedia Ontology Classification Dataset | | Dataset | DBPedia Ontology Classification Dataset |
| Training Parameters | epoch=5, batch_size=128 | | Training Parameters | epoch=5, batch_size=4096 |
| Optimizer | Adam | | Optimizer | Adam |
| Loss Function | Softmax Cross Entropy | | Loss Function | Softmax Cross Entropy |
| outputs | probability | | outputs | probability |
| Speed | 60ms/step (8pcs) | | Speed | 58ms/step (1pcs) |
| Total Time | 164s (8pcs) | | Epoch Time | 8.15s (1pcs) |
| Loss | 2.6e-5 | | Loss | 2.6e-4 |
| Params (M) | 106 | | Params (M) | 106 |
| Checkpoint for inference | 1.2G (.ckpt file) | | Checkpoint for inference | 1.2G (.ckpt file) |
| Scripts | [fasttext](https://gitee.com/mindspore/mindspore/tree/master/model_zoo/official/nlp/fasttext) | | Scripts | [fasttext](https://gitee.com/mindspore/mindspore/tree/master/model_zoo/official/nlp/fasttext) |
@ -201,13 +202,13 @@ Parameters for both training and evaluation can be set in config.py. All the dat
| uploaded Date | 11/21/2020 (month/day/year) | | uploaded Date | 11/21/2020 (month/day/year) |
| MindSpore Version | 1.1.0 | | MindSpore Version | 1.1.0 |
| Dataset | Yelp Review Polarity Dataset | | Dataset | Yelp Review Polarity Dataset |
| Training Parameters | epoch=5, batch_size=128 | | Training Parameters | epoch=5, batch_size=2048 |
| Optimizer | Adam | | Optimizer | Adam |
| Loss Function | Softmax Cross Entropy | | Loss Function | Softmax Cross Entropy |
| outputs | probability | | outputs | probability |
| Speed | 74ms/step (8pcs) | | Speed | 101ms/step (1pcs) |
| Total Time | 195s (8pcs) | | Epoch Time | 28s (1pcs) |
| Loss | 7.7e-4 | | Loss | 0.062 |
| Params (M) | 103 | | Params (M) | 103 |
| Checkpoint for inference | 1.2G (.ckpt file) | | Checkpoint for inference | 1.2G (.ckpt file) |
| Scripts | [fasttext](https://gitee.com/mindspore/mindspore/tree/master/model_zoo/official/nlp/fasttext) | | Scripts | [fasttext](https://gitee.com/mindspore/mindspore/tree/master/model_zoo/official/nlp/fasttext) |
@ -220,8 +221,8 @@ Parameters for both training and evaluation can be set in config.py. All the dat
| Uploaded Date | 12/21/2020 (month/day/year) | | Uploaded Date | 12/21/2020 (month/day/year) |
| MindSpore Version | 1.1.0 | | MindSpore Version | 1.1.0 |
| Dataset | AG's News Topic Classification Dataset | | Dataset | AG's News Topic Classification Dataset |
| batch_size | 128 | | batch_size | 512 |
| Total Time | 66s | | Epoch Time | 2.36s |
| outputs | label index | | outputs | label index |
| Accuracy | 92.53 | | Accuracy | 92.53 |
| Model for inference | 254M (.ckpt file) | | Model for inference | 254M (.ckpt file) |
@ -232,8 +233,8 @@ Parameters for both training and evaluation can be set in config.py. All the dat
| Uploaded Date | 12/21/2020 (month/day/year) | | Uploaded Date | 12/21/2020 (month/day/year) |
| MindSpore Version | 1.1.0 | | MindSpore Version | 1.1.0 |
| Dataset | DBPedia Ontology Classification Dataset | | Dataset | DBPedia Ontology Classification Dataset |
| batch_size | 128 | | batch_size | 4096 |
| Total Time | 164s | | Epoch Time | 8.15s |
| outputs | label index | | outputs | label index |
| Accuracy | 98.6 | | Accuracy | 98.6 |
| Model for inference | 1.2G (.ckpt file) | | Model for inference | 1.2G (.ckpt file) |
@ -244,8 +245,8 @@ Parameters for both training and evaluation can be set in config.py. All the dat
| Uploaded Date | 12/21/2020 (month/day/year) | | Uploaded Date | 12/21/2020 (month/day/year) |
| MindSpore Version | 1.1.0 | | MindSpore Version | 1.1.0 |
| Dataset | Yelp Review Polarity Dataset | | Dataset | Yelp Review Polarity Dataset |
| batch_size | 128 | | batch_size | 2048 |
| Total Time | 195s | | Epoch Time | 28s |
| outputs | label index | | outputs | label index |
| Accuracy | 95.7 | | Accuracy | 95.7 |
| Model for inference | 1.2G (.ckpt file) | | Model for inference | 1.2G (.ckpt file) |

1
model_zoo/official/nlp/fasttext/scripts/create_dataset.sh
View File

@ -32,7 +32,6 @@ SOURCE_DATASET_PATH=$(get_real_path $1)
DATASET_NAME=$2 DATASET_NAME=$2
export DEVICE_NUM=1 export DEVICE_NUM=1
export DEVICE_ID=5
export RANK_ID=0 export RANK_ID=0
export RANK_SIZE=1 export RANK_SIZE=1

3
model_zoo/official/nlp/fasttext/scripts/run_eval.sh
View File

@ -32,8 +32,9 @@ DATASET=$(get_real_path $1)
echo $DATASET echo $DATASET
DATANAME=$2 DATANAME=$2
MODEL_CKPT=$(get_real_path $3) MODEL_CKPT=$(get_real_path $3)
DEVICEID=$4
export DEVICE_NUM=1 export DEVICE_NUM=1
export DEVICE_ID=5 export DEVICE_ID=$DEVICEID
export RANK_ID=0 export RANK_ID=0
export RANK_SIZE=1 export RANK_SIZE=1

3
model_zoo/official/nlp/fasttext/scripts/run_standalone_train.sh
View File

@ -32,9 +32,10 @@ DATASET=$(get_real_path $1)
echo $DATASET echo $DATASET
DATANAME=$(basename $DATASET) DATANAME=$(basename $DATASET)
echo $DATANAME echo $DATANAME
DEVICEID=$2
export DEVICE_NUM=1 export DEVICE_NUM=1
export DEVICE_ID=0 export DEVICE_ID=$DEVICEID
export RANK_ID=0 export RANK_ID=0
export RANK_SIZE=1 export RANK_SIZE=1

4
model_zoo/official/nlp/fasttext/train.py
View File

@ -141,12 +141,12 @@ def _build_training_pipeline(pre_dataset):
callbacks = [time_monitor, loss_monitor] callbacks = [time_monitor, loss_monitor]
if rank_size is None or int(rank_size) == 1: if rank_size is None or int(rank_size) == 1:
ckpt_callback = ModelCheckpoint(prefix='fasttext', ckpt_callback = ModelCheckpoint(prefix='fasttext',
directory=os.path.join('./', 'ckpe_{}'.format(os.getenv("DEVICE_ID"))), directory=os.path.join('./', 'ckpt_{}'.format(os.getenv("DEVICE_ID"))),
config=ckpt_config) config=ckpt_config)
callbacks.append(ckpt_callback) callbacks.append(ckpt_callback)
if rank_size is not None and int(rank_size) > 1 and MultiAscend.get_rank() % 8 == 0: if rank_size is not None and int(rank_size) > 1 and MultiAscend.get_rank() % 8 == 0:
ckpt_callback = ModelCheckpoint(prefix='fasttext', ckpt_callback = ModelCheckpoint(prefix='fasttext',
directory=os.path.join('./', 'ckpe_{}'.format(os.getenv("DEVICE_ID"))), directory=os.path.join('./', 'ckpt_{}'.format(os.getenv("DEVICE_ID"))),
config=ckpt_config) config=ckpt_config)
callbacks.append(ckpt_callback) callbacks.append(ckpt_callback)
print("Prepare to Training....") print("Prepare to Training....")