| Add conversion script | 3 年前 |
| blank lm | 5 年前 |
| Fix config for 10B model | 5 年前 |
| fix df_finetune_seq2seq.sh save path --save should use SAVE_PATH not CHECKPOINT_PATH | 3 年前 |
| add task_Clue-WSC and task_CMRC | 4 年前 |
| Update config | 5 年前 |
| Update config and script | 5 年前 |
| Merge branch 'release' into 10B-release # Conflicts: # README.md # config/config_blockta_large.json # config/config_gpt_10B.json # config/ds_block_large.sh # config/ds_blockta_large.sh # config_tasks/model_blocklm_large.sh # config_tasks/model_blocklm_roberta_large.sh # config_tasks/seq_cnndm.sh # config_tasks/seq_cnndm_org.sh # config_tasks/zero_lambada.sh # config_tasks/zero_lambada_uni.sh # config_tasks/zero_wikitext.sh # docker/cuda102.dockerfile # finetune_glm.py # generate_samples.py # model/downstream.py # model/modeling_glm.py # pretrain_glm.py # process_grid.py # scripts/ds_pretrain_nvidia.sh # scripts/evaluate_lm.sh # scripts/finetune_seq2seq.sh # scripts/finetune_superglue.sh # scripts/finetune_superglue_grid.sh # scripts/generate_block.sh # tasks/seq2seq/finetune.py # tasks/superglue/dataset.py # tasks/superglue/finetune.py # tasks/superglue/pvp.py # train_utils.py | 5 年前 |
| Add zero-shot multiple choice | 3 年前 |
| Update rouge evaluation | 5 年前 |
| blank language model | 5 年前 |
| implement SQuAD (cherry picked from commit 2b7f038d2326e6b7f1a46b99459ef6870df3bdc7) | 4 年前 |
| implement SQuAD (cherry picked from commit 2b7f038d2326e6b7f1a46b99459ef6870df3bdc7) | 4 年前 |
| implement SQuAD (cherry picked from commit 2b7f038d2326e6b7f1a46b99459ef6870df3bdc7) | 4 年前 |
| rename to GLM; add param description | 5 年前 |
| Merge changes from mp_finetune | 5 年前 |
| Add top_p and top_k for seq2seq evaluation | 3 年前 |
| blank lm | 5 年前 |