GLM is a General Language Model pretrained with an autoregressive blank-filling objective and can be finetuned on various natural language understanding and generation tasks.

Please refer to our paper for a detailed description of GLM:

All NLP Tasks Are Generation Tasks: A General Pretraining Framework

Zhengxiao Du*, Yujie Qian*, Xiao Liu, Ming Ding, Jiezhong Qiu, Zhilin Yang, Jie Tang (*: equal contribution)

Part of the code is based on Megatron-LM and PET.

You can download the pretrained models used in the paper here.

Clone this repo

git clone https://github.com/THUDM/GLM
cd GLM

Please first install PyTorch (we use 1.7.0) and apex, and then install other dependencies by

pip install -r requirements.txt

We provide scripts for finetuning GLM on some downstream tasks.

• Download the SuperGlue data and check the experiment setup in scripts/finetune_superglue.sh. Note that DATA_ROOT, CHECKPOINT_PATH, SAVE_PATH need to be changed to your local path. You may also change the batch-size and nproc_per_node according to your available hardware. We suggest to use aggregated batch size 64 for MultiRC and ReCORD and 16 for other tasks.

• Run the following script (use the COPA dataset as an example)

bash scripts/finetune_superglue.sh \
     config_tasks/model_blocklm_roberta_large.sh \
     config_tasks/task_copa.sh

• To apply GLM to a new NLU dataset with cloze-filling finetuning, implement a DataProcessor in tasks/superglue/dataset.py for data loading and add a PVP in tasks/superglue/pvp.py for the cloze question. More details can be found here.

• The cloze questions (prompts) used in this work are written by human. We are also studying a P-tuning (prompt tuning) approach to search for the optimal continuous prompt. Please refer to our paper and code.

• Download the Gigaword dataset and check the experiment setup in scripts/finetune_seq2seq.sh. Change DATA_ROOT, CHECKPOINT_PATH, SAVE_PATH to your local path.

• Run the following script

bash scripts/finetune_seq2seq.sh \ 
     config_tasks/model_blocklm_large_generation.sh \ 
     config_tasks/seq_gigaword.sh

• For calculating rouge, install file2rouge from here and run bash scripts/evaluate_seq2seq.sh

• Download the LAMBADA data and change DATA_ROOT, CHECKPOINT_PATH in scripts/evaluate_lm.sh
• Run the following script

bash scripts/evaluate_lm.sh \ 
     config_tasks/model_blocklm_large_generation.sh \
     config_tasks/zero_lambada.sh 

• Download our test set of wikibook (or any dataset following the same format) and change DATA_ROOT, CHECKPOINT_PATH in scripts/evaluate_lm.sh
• Run the following script

  bash scripts/evaluate_lm.sh \ 
     config_tasks/model_blocklm_large_generation.sh \
     config_tasks/zero_lm.sh 

• Download the Yahoo dataset and check the experiment setup in scripts/finetune_blank.sh. Change DATA_ROOT, CHECKPOINT_PATH, SAVE_PATH to your local path.

• Run the following script

bash scripts/finetune_blank.sh \ 
     config_tasks/model_blocklm_large.sh \ 
     config_tasks/seq_blank.sh

• Change CHECKPOINT_PATH to your local path. Run the following script

bash scripts/generate_block.sh \
     config_tasks/model_blocklm_large.sh

Example:

Context: Ng is an adjunct professor at [MASK] (formerly associate professor and Director of its Stanford AI Lab or SAIL ). Also a pioneer in online education, Ng co-founded Coursera and deeplearning.ai.

GLM: [CLS] ng is an adjunct professor at [MASK] ( formerly associate professor and director of its stanford ai lab or sail ) . also a pioneer in online education , ng co - founded coursera and deeplearning . ai . [PAD] <|startofpiece|> the stanford university

Please cite our paper if you find this code useful for your research:

@article{DBLP:journals/corr/abs-2103-10360,
  author    = {Zhengxiao Du and
               Yujie Qian and
               Xiao Liu and
               Ming Ding and
               Jiezhong Qiu and
               Zhilin Yang and
               Jie Tang},
  title     = {All {NLP} Tasks Are Generation Tasks: {A} General Pretraining Framework},
  journal   = {CoRR},
  volume    = {abs/2103.10360},
  year      = {2021},
  url       = {https://arxiv.org/abs/2103.10360}
}