Skip to content
Snippets Groups Projects
Commit 4a6880b5 authored by i-robot's avatar i-robot Committed by Gitee
Browse files

!162 fix warpctc train failed issue 

Merge pull request !162 from gengdongjie/fix_issues_master
parents b2846f73 4f1e6874
No related branches found
No related tags found
No related merge requests found
Hide whitespace changes
Inline Side-by-side
official/cv/warpctc/src/warpctc_for_train.py
+ 1
− 2
View file @ 4a6880b5
......@@ -105,5 +105,4 @@ class TrainOneStepCellWithGradClip(Cell):
if self.reducer_flag:
# apply grad reducer on grads
grads = self.grad_reducer(grads)
self.optimizer(grads)
return loss
return F.depend(loss, self.optimizer(grads))
research/nlp/ktnet/README_CN.md
+ 28
− 28
View file @ 4a6880b5
......@@ -269,7 +269,7 @@ mv wn_concept2vec.txt nell_concept2vec.txt data/KB_embeddings
--checkpoints output
```
- record数据集的需要训练步骤把‘python scripts/run_KTNET_squad.py’换为‘python scripts/run_KTNET_record.py’
- record数据集的需要训练步骤把 `python scripts/run_KTNET_squad.py` 换为 `python scripts/run_KTNET_record.py`
### 评估
......@@ -313,7 +313,7 @@ mv wn_concept2vec.txt nell_concept2vec.txt data/KB_embeddings
--checkpoints
```
- record数据集的需要评估步骤把‘bash scripts/run_KTNET_squad_eval.sh’换为‘bash scripts/run_KTNET_record_eval.sh’
- record数据集的需要评估步骤把 `bash scripts/run_KTNET_squad_eval.sh` 换为 `bash scripts/run_KTNET_record_eval.sh`
## 训练过程
......@@ -325,7 +325,7 @@ mv wn_concept2vec.txt nell_concept2vec.txt data/KB_embeddings
# 单机
bash scripts/run_squad_twomemory.sh [DATAPATH]
# 分布式训练 8卡
bash scripts/run_squad_twomemory.sh [DATAPATH] [RANK_TABLE_FILE]
bash run_squad_twomemory_distribute.sh [DATAPATH] [RANK_TABLE_FILE]
```
DATAPATH为必选项，为数据文件存放的路径。
......@@ -349,7 +349,7 @@ python run_KTNET_squad.py
# 单机
bash scripts/run_record_twomemory.sh [DATAPATH]
# 分布式训练 8卡
bash scripts/run_record_twomemory.sh [DATAPATH] [RANK_TABLE_FILE]
bash run_record_twomemory_distribute.sh [DATAPATH] [RANK_TABLE_FILE]
```
DATAPATH为必选项，为数据文件存放的路径。
......@@ -438,33 +438,33 @@ NEED_PREPROCESS为必选项, 在[y|n]中取值，表示数据是否预处理为b
### 训练性能
| 参数 | Ascend |Ascend |
| --------------------------| ---------------------------------------------------------- |---------------------------------------------------------- |
| 模型 | KTNET |KTNET |
| 资源 |Ascend 910；CPU 2.60GHz，192核；内存 755GB；系统 Euler2.8 |Ascend 910；CPU 2.60GHz，192核；内存 755GB；系统 Euler2.8 |
| 上传日期 | 2021-05-12 | 2021-05-12 |
| 数据集 |ReCoRD |SQuAD |
| 训练参数 | src/config.py | src/config.py |
| 学习率 | 7e-5 | 4e-5 |
| 优化器 | Adam | Adam |
| 损失函数 | SoftmaxCrossEntropy | SoftmaxCrossEntropy |
| 轮次 | 4 | 3 |
| Batch_size | 12*8 | 8*8 |
| 损失 |0.31248128 |0.35267675 |
| 速度 | 428毫秒/步 | 338毫秒/步 |
| 总时长 | 2.5小时 | 1小时 |
| 参数 | Ascend |Ascend |
| -------------| ---------------------------------------------------------- |---------------------------------------------------------- |
| 模型 | KTNET | KTNET |
| 资源 | Ascend 910；CPU 2.60GHz，192核；内存 755GB；系统 Euler2.8 | Ascend 910；CPU 2.60GHz，192核；内存 755GB；系统 Euler2.8 |
| 上传日期 | 2021-05-12 | 2021-05-12 |
| 数据集 | ReCoRD | SQuAD |
| 训练参数 | src/config.py | src/config.py |
| 学习率 | 7e-5 | 4e-5 |
| 优化器 | Adam | Adam |
| 损失函数 | SoftmaxCrossEntropy | SoftmaxCrossEntropy |
| 轮次 | 4 | 3 |
| Batch_size | 12*8 | 8*8 |
| 损失 | 0.31248128 | 0.35267675 |
| 速度 | 428毫秒/步 | 338毫秒/步 |
| 总时长 | 2.5小时 | 1小时 |
### 推理性能
| 参数 | Ascend | Ascend |
| -------------------------- | ----------------------------- | ----------------------------- |
| 模型 | KTNET | KTNET |
| 数据集 |ReCoRD |ReCoRD |
| 上传日期 |2021-05-12 |2021-05-12 |
| 数据集 | ReCoRD | ReCoRD |
| f1 | 71.48 | 91.31 |
| exact_match | 69.61 | 84.38 |
| 总时长 | 15分钟 | 15分钟 |
| 参数 | Ascend | Ascend |
| ----------------| ---------------| --------------|
| 模型 | KTNET | KTNET |
| 数据集 | ReCoRD | ReCoRD |
| 上传日期 | 2021-05-12 | 2021-05-12 |
| 数据集 | ReCoRD | ReCoRD |
| f1 | 71.48 | 91.31 |
| exact_match | 69.61 | 84.38 |
| 总时长 | 15分钟 | 15分钟 |
# ModelZoo主页
......
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment