Skip to content

Latest commit

 

History

History
52 lines (36 loc) · 988 Bytes

README.md

File metadata and controls

52 lines (36 loc) · 988 Bytes
Raw

Seq2Seq

This is a pytorch implementation of Seq2Seq model for dialogue generation.

Requirements

  • Python>=3.6
  • PyTorch>=1.0
  • tqdm
  • numpy
  • nltk
  • scikit-learn

Dataset

The dataset can be downloaded from 2019 Language and Intelligence Challenge, we just put example train data under the folder ./data/resource.

Quickstart

Step 1: Preprocess the data

Put the data provided by the organizer under the data folder and rename them train/dev/test.txt:

./data/resource/train.txt
./data/resource/dev.txt
./data/resource/test.txt

Process the data with the following commands:

bash process_data.sh

Step 2: Train the model

Train model with the following commands:

bash run_train.sh

Step 3: Test the model

Test model with the following commands:

bash run_test.sh

References

[1] https://github.com/baidu/knowledge-driven-dialogue/tree/master/generative_pt