SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

Add Mega: Moving Average Equipped Gated Attention (#21766)

Browse Source 
* add mega file structure and plain pytorch version of mega source code

* added config class with old naming conventions

* filled in mega documentation

* added config class and embeddings with optional token types

* updated notes

* starting the conversion process, deleted intermediate and added use_cache back to config

* renamed config attributes in modeling_mega.py

* checkpointing before refactoring incremental decoding functions

* removed stateful incremental key/values for EMA and self-attention

* refactored MovingAverageGatedAttention to remove stateful k/v history and use unified attention mask

* MovingAverageGatedAttention works with incremental decoding + past values, added sequence length enforcement

* more comments in MovingAverageGatedAttention + checkpointing before GatedCrossAttention

* bug fix in attention mask handling in MovingAverageGatedAttention

* removed incremental state from GatedCrossAttention and removed IncrementalState class

* finished gated cross attention and got MegaLayer working

* fixed causal masking in mega decoder

* fixed how padding and causal masks are passed through MegaLayer with and without k/v caching

* finished MegaModel; tested with encoder, decoder-only, and cross-attention type inputs; started work on downstream classes; removed mentions of position_ids

* added optional dense hidden layer for masked and causal LM classes

* docstring updates in MultiHeadEMA and GatedCrossAttention, removed unnecessary inputs in cross-attention

* removed before_attn_fn in Mega class and updated docstrings and comments up to there

* bug fix in MovingAverageGatedAttention masking

* working conversion of MLM checkpoint in scratchpad script -- perfect matches

* moved arg for hidden dense layer in LM head to config; discovered issue where from_pretrained is renaming gamma and beta parameters

* renamed gamma and beta parameters to avoid HF renaming when loading from checkpoint

* finished checkpoint conversion script

* cleanup old class in mega config script

* removed 'copied from' statements and passing integration tests

* added num_attention_heads=1 to config for integration compatibility, decoder tests working, generation tests failing

* fixed tuple output of megamodel

* all common tests passing after fixing issues in decoder, gradient retention, and initialization

* added mega-specific tests, ready for more documentation and style checks

* updated docstrings; checkpoint before style fixes

* style and quality checks, fixed initialization problem in float_tensor, ready for PR

* added mega to toctree

* removed unnecessary arg in megaconfig

* removed unused arg and fixed code samples with leftover roberta models

* Apply suggestions from code review

Applied all suggestions except the one renaming a class, as I'll need to update that througout

Co-authored-by: Arthur <48595927+ArthurZucker@users.noreply.github.com>

* fixed issue where .view breaks batch dimension, conversion script fixed with absolute imports, updated readme with Mega->MEGA

* removed asserts in Mega code, renamed sequencenorm, gatedcrossattention, and NFFN, replaced get_activation_fn with ACTFN, and added sequencenorm to layer norms

* reformatted .forward() docstrings to match style and removed unused mask input in cross-attention

* removed all reset_parameters() methods and rolled into MegaPreTrainedModel._init_weights()

* renamed all single-letter variables and improved readability in tensor size comments, Mega->MEGA in 2 documentation files

* variable names in NFFN

* manual Mega->MEGA changes in docs

* Mega->MEGA in config auto

* style and quality fixes

* Apply suggestions from code review

Co-authored-by: Arthur <48595927+ArthurZucker@users.noreply.github.com>

* renamed parameters and variables with confusing names, added copied from statements, moved fft conv to its own method, other cleanup from PR comments

* commit before dealing with merge conflicts

* made new attention activation functions available in ACT2FN and added generation test from OPT

* style and quality in activations and tests

* documentation fixes, renaming variables in dropout and rotary positions, used built-in causal masking, encoders->layers in MegaModel, moved comments into docstrings

* style and quality fixes after latest updates, before rotary position ids

* causal mask in MegaBlock docstring + added missing device passing

* Apply suggestions from code review

Co-authored-by: Arthur <48595927+ArthurZucker@users.noreply.github.com>

* Update README.md

Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

* added Mega prefixes where missing, reverted MegaSequenceNorm to if-else, other module renaming requested in PR

* style and quality fixes + readme updates pointing to main

---------

Co-authored-by: Arthur <48595927+ArthurZucker@users.noreply.github.com>
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
This commit is contained in:
Mitch Naylor
2023-03-24 07:17:27 -05:00
committed by GitHub
parent 0fa46524ac
commit 57f25f4b7f
30 changed files with 3790 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
README_ko.md
View File

@@ -314,6 +314,7 @@ Flax, PyTorch, TensorFlow 설치 페이지에서 이들을 conda로 설치하는
1. **[MaskFormer](https://huggingface.co/docs/transformers/model_doc/maskformer)** (Meta and UIUC 에서) Bowen Cheng, Alexander G. Schwing, Alexander Kirillov 의 [Per-Pixel Classification is Not All You Need for Semantic Segmentation](https://arxiv.org/abs/2107.06278) 논문과 함께 발표했습니다.
1. **[mBART](https://huggingface.co/docs/transformers/model_doc/mbart)** (Facebook 에서) Yinhan Liu, Jiatao Gu, Naman Goyal, Xian Li, Sergey Edunov, Marjan Ghazvininejad, Mike Lewis, Luke Zettlemoyer 의 [Multilingual Denoising Pre-training for Neural Machine Translation](https://arxiv.org/abs/2001.08210) 논문과 함께 발표했습니다.
1. **[mBART-50](https://huggingface.co/docs/transformers/model_doc/mbart)** (Facebook 에서) Yuqing Tang, Chau Tran, Xian Li, Peng-Jen Chen, Naman Goyal, Vishrav Chaudhary, Jiatao Gu, Angela Fan 의 [Multilingual Translation with Extensible Multilingual Pretraining and Finetuning](https://arxiv.org/abs/2008.00401) 논문과 함께 발표했습니다.
1. **[MEGA](https://huggingface.co/docs/transformers/main/model_doc/mega)** (Facebook 에서 제공)은 Xuezhe Ma, Chunting Zhou, Xiang Kong, Junxian He, Liangke Gui, Graham Neubig, Jonathan May, and Luke Zettlemoyer.의 [Mega: Moving Average Equipped Gated Attention](https://arxiv.org/abs/2209.10655)논문과 함께 발표했습니다.
1. **[Megatron-BERT](https://huggingface.co/docs/transformers/model_doc/megatron-bert)** (NVIDIA 에서) Mohammad Shoeybi, Mostofa Patwary, Raul Puri, Patrick LeGresley, Jared Casper and Bryan Catanzaro 의 [Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism](https://arxiv.org/abs/1909.08053) 논문과 함께 발표했습니다.
1. **[Megatron-GPT2](https://huggingface.co/docs/transformers/model_doc/megatron_gpt2)** (NVIDIA 에서) Mohammad Shoeybi, Mostofa Patwary, Raul Puri, Patrick LeGresley, Jared Casper and Bryan Catanzaro 의 [Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism](https://arxiv.org/abs/1909.08053) 논문과 함께 발표했습니다.
1. **[MGP-STR](https://huggingface.co/docs/transformers/model_doc/mgp-str)** (Alibaba Research 에서 제공)은 Peng Wang, Cheng Da, and Cong Yao.의 [Multi-Granularity Prediction for Scene Text Recognition](https://arxiv.org/abs/2209.03592)논문과 함께 발표했습니다.