Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Dev liuyibo #70

Open
wants to merge 4 commits into
base: main
Choose a base branch
from
Open

Dev liuyibo #70

wants to merge 4 commits into from

Conversation

rookielyb
Copy link
Contributor

1.增加Keyphrase Generation项目
2.在examples/text_generation/ 新增 Keyphrase Generation的启动项 以及少量数据
3.新增gts_engine/pipelines/qiankunding_generation.py 主代码
3.新增gts_engine/qiankunding/dataloaders/text_generation/dataloader_kgt5.py的数据处理代码
4.新增gts_engine/qiankunding/models/text_generation/t5_kg.py 模型代码
5.在gts_engine/qiankunding/utils/evaluation.py 中新增 TextGenerateEvaluator 类 用于Generate的评估
6.在gts_engine/qiankunding/utils/tokenization.py中新增 T5的tokenization
7.运行examples/text_generation/run_train_qiankunding.sh 成功如下
企业微信截图_6115ece4-cd12-42f3-bb03-2287ae72ec5b

8.运行examples/text_generation/run_inference_qiankunding.sh 成功如下
企业微信截图_928e3f9a-dbc0-4fe5-ac39-bdc49ec50a8c

@pskun pskun self-requested a review December 30, 2022 06:36
python gts_engine/gts_engine_inference.py \
--task_dir=$TASK_DIR \
--engine_type=qiankunding \
--task_type=generation \
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

task type改成keyphrase_generation

--engine_type=qiankunding \
--train_mode=standard \
--task_dir=$TASK_DIR \
--task_type=generation \
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

task type改成keyphrase_generation

--data_dir=$WORK_DIR/examples/text_generation \
--save_path=$TASK_DIR/outputs \
--pretrained_model_dir=$PRETRAINED_DIR \
--train_batchsize=32 \
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

测过这么大的bs占多少显存吗

@@ -0,0 +1,141 @@
import json
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

目录改成:gts_engine/qiankunding/dataloaders/keyphrase_generation

@@ -0,0 +1,181 @@
from genericpath import exists
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

改成:gts_engine/qiankunding/models/keyphrase_generation

logger = Logger().get_log()


class T5KG(BaseModel):
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

叫KeyphraseGenerationT5吧,本身单词也不长

inputs = self.train_inputs(batch)
outputs = self.model.generate(
input_ids = inputs['input_ids'],
max_length = 32,
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

这个32是拍的?


outputs = self.model.generate(
input_ids = inputs['input_ids'],
max_length=32
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

32这个,设置成类成员变量吧


results.append(pred)
TP, total_pred, total_true = 0, 0, 0
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

这里的计算和前面模型的validation_step差不多,为啥不单独抽象出一个公共的函数

special_tokens += [f"[choice{i+1}]" for i in range(200)]
# special_tokens += [f"{i+1}" for i in range(200)]

print("pretrained_model_path", pretrained_model_path)
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

没用的输出要删掉

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants