Commits · 308e87b02bbe27c83883272de2ec9199880d0fd3 · mirrored_repos / MachineLearning / thukeg / SwissArmyTransformer

This project is mirrored from https://github.com/THUDM/SwissArmyTransformer. Pull mirroring failed 2 months ago.
Repository mirroring has been paused due to too many failed attempts. It can be resumed by a project maintainer or owner.
Last successful update 2 months ago.

Dec 12, 2021
- Add support for T5 v1.1 · 308e87b0
  duzx16 authored 3 years ago
  
  308e87b0
- Better test script for transformers t5 · 57d82dff
  duzx16 authored 3 years ago
  
  57d82dff
Dec 10, 2021
- Update implementation for T5 · 913ce150
  duzx16 authored 3 years ago
  
  913ce150
- Update script for backward test · fdcf4735
  duzx16 authored 3 years ago
  
  fdcf4735
- Update script for backward test · adf4133f
  duzx16 authored 3 years ago
  
  adf4133f
Dec 05, 2021
- remove layer_forward enc-dec params for simple · bd2d86f9
  Ming Ding authored 3 years ago
  
  bd2d86f9
- Add T5 Tokenizer · 049b6ffc
  duzx16 authored 3 years ago
  
  049b6ffc
- Add default attention mask for EncoderDecoder · 3fc1af82
  duzx16 authored 3 years ago
  
  Remove position id arguments for T5
  3fc1af82
- Fix bug in Transformer implementation · 63c5ce73
  duzx16 authored 3 years ago
  
  63c5ce73
- Merge remote-tracking branch 'origin/enc-dec' into enc-dec · 07d5dc0f
  duzx16 authored 3 years ago
  
  07d5dc0f
- Add test scripts for T5 · f8e4c9f0
  duzx16 authored 3 years ago
  
  f8e4c9f0
- tidy up enc-dec model · 6a2fa408
  Ming Ding authored 3 years ago
  
  6a2fa408
Dec 04, 2021
- Merge branch 'tmp_eval_support' into enc-dec · 3ab8731b
  Ming Ding authored 3 years ago
  
  3ab8731b
- add strict-eval support · 44333e0f
  Ming Ding authored 3 years ago
  
  44333e0f
- Add requires_grad at the beginning · 00442c04
  duzx16 authored 3 years ago
  
  00442c04
- Remove common layers · be349db4
  duzx16 authored 3 years ago
  
  be349db4
- Fix scaling_attention_score in standard_attention · 0ac85141
  duzx16 authored 3 years ago
  
  0ac85141
- Fix T5 Model · 9bedc566
  duzx16 authored 3 years ago
  
  9bedc566
Dec 02, 2021
- Implement T5 model · cc852e00
  duzx16 authored 3 years ago
  
  cc852e00
Dec 01, 2021
- Reformat code · 7aee8e24
  duzx16 authored 3 years ago
  
  7aee8e24
- Reformat code · c761af73
  duzx16 authored 3 years ago
  
  c761af73
- Merge branch 'package' of github.com:THUDM/SwissArmyTransformer into package · b32f9dde
  duzx16 authored 3 years ago
  
  Conflicts: SwissArmyTransformer/mpu/transformer.py SwissArmyTransformer/training/deepspeed_training.py
  b32f9dde
- Accept config_params for deepspeed engine · b84c5aa0
  duzx16 authored 3 years ago
  
  b84c5aa0
Nov 30, 2021
- Merge branch 'package' of github.com:THUDM/SwissArmyTransformer into package · 33bbb65c
  Ming Ding authored 3 years ago
  
  33bbb65c
- add tensorboard log of customized logs · 9eca6ab3
  Ming Ding authored 3 years ago
  
  9eca6ab3
Nov 26, 2021
- Update README.md · cbe869e4
  ledw-2 authored 3 years ago
  
  Add more detailed example in section Run GLM
  cbe869e4
Nov 25, 2021
- fix typo encdec · 98a5247a
  Ming Ding authored 3 years ago
  
  98a5247a
- fix pass encdec args · 55884dae
  Ming Ding authored 3 years ago
  
  55884dae
- fix shared crossattn bug · 2721c2fa
  Ming Ding authored 3 years ago
  
  2721c2fa
Nov 23, 2021
- Merge branch 'package' of github.com:THUDM/SwissArmyTransformer into package · 233d9495
  Ming Ding authored 3 years ago
  
  233d9495
- add untest enc-dec and fix email · f191e5be
  Ming Ding authored 3 years ago
  
  f191e5be
- Add keyword arguments for forward_step · 49dea453
  Zhengxiao Du authored 3 years ago
  
  49dea453
- Add output_hidden_states in BaseTransformer · 97652f24
  Zhengxiao Du authored 3 years ago
  
  97652f24
Nov 21, 2021
- Fix report memory in utils.py · d554d437
  Zhengxiao Du authored 3 years ago
  
  d554d437
Nov 14, 2021
- Add del_mixin function · f455c777
  Zhengxiao Du authored 3 years ago
  
  f455c777
Nov 09, 2021
- Export master ip and port to args · 91372f1c
  Zhengxiao Du authored 3 years ago
  
  91372f1c
Nov 08, 2021
- Add parallel_output argument for GLMModel · 61ef4c7c
  Zhengxiao Du authored 3 years ago
  
  61ef4c7c
- Add parallel_output argument for GLMModel · b4eaba16
  Zhengxiao Du authored 3 years ago
  
  b4eaba16
Nov 07, 2021
- Fix parallel_output and add argument · 826b7d53
  Zhengxiao Du authored 3 years ago
  
  826b7d53
Nov 06, 2021
- Fix train_step while gradient accumulation · aff0493f
  Zhengxiao Du authored 3 years ago
  
  aff0493f