The RWKV Language Model

Related tags

Text Data & NLPRWKV-LM
Overview

RWKV-LM

We propose the RWKV language model, with alternating time-mix and channel-mix layers:

\begin{align*} \text{Time-mix :} && \text{TM}_{t,c} &&=&&\text{sigmoid}(\text{R}_{t,c}) &&\cdot&& &&\textstyle\sum_{u} &&\textbf{W}_{t,u,c} &&\cdot&& \text{softmax}_t(\text{K}_{u,c}) &&\cdot&& \text{V}_{u,c}\\ \text{Channel-mix :} && \text{CM}_{t,c} &&=&&\text{sigmoid}(\text{R}_{t,c}) &&\cdot&& &&\textstyle\sum_d &&\textbf{W}_{c,d} &&\cdot&& \text{gelu}(\text{K}_{t,d}) &&\cdot&& \text{V}_{t,d} \end{align*}

  • The R, K, V are generated by linear transforms of input, and W is parameter. The idea of RWKV is to decompose attention into R(target) * W(src, target) * K(src). So we can call R "receptance", and sigmoid means it's in 0~1 range.

  • The Time-mix is similar to AFT (https://arxiv.org/abs/2105.14103). There are two differences.

(1) We changed the normalization (denominator). For masked language models, we define:

\text{softmax}_t(\text{K}_{u,c}) = \frac{\exp(\text{K}_{u,c})}{\sum_{v \leq t}\exp(\text{K}_{v,c})}

(2) We decompose W_{t,u,c} and introduce multi-head W (here h is the corresponding head of c):

W_{t,u,c}=f_h(t-u)\cdot \alpha_h(u) \cdot \beta_h(t)

(3) You don't need LayerNorm for Time-mix. In fact, the model converges faster when LayerNorm is removed.

Moreover we multiply the final output of Time-mix layer by γ(t). The reason for the α β γ factors, is because the context size is smaller when t is small, and this can be compensated using the α β γ factors.

We also propose a new sampling method (as in src/utils.py):

(1) Find the max probability p_max after softmax.

(2) Remove all entries whose probability is lower than 0.02 * pow(p_max, 2)

(3) Feel free to tune the 0.02 and 2 factor.

Training loss, RWKV vs MHA+Rotary+GeGLU:

RWKV-vs-MHA

(this is character-level loss with simplebooks-92 dataset https://dldata-public.s3.us-east-2.amazonaws.com/simplebooks.zip)

Comments
  • Sequence to Sequence?

    Sequence to Sequence?

    Hey @BlinkDL! Awesome project!

    I was wondering if you have performed any Seq-2-Seq experiments with it? Any reason for going with GPT model in the first place as opposed to something like T5 (standard Transformer)? Any direction on what changes will be required to make a standard encoder-decoder architecture with RWKV?

    Also, is there any report on in-context-learning/FSL capability of the latest trained model?

    opened by SushantDaga 2
  • v4 model.py vs model_run.py

    v4 model.py vs model_run.py

    Hi, Thanks for this awesome repo! I'm trying to understand the code and found that in the v4 folder, there's this model.py and model_run.py, which contains GPT and RWKV_GPT respectively which all uses different initialization methods. Could you elaborate on when should which one be used? Thanks in advnace!

    opened by jingweiz 3
  • RWKV-4 169m/430m in browser with ORT Web / TF.js / tfjs-tflite?

    RWKV-4 169m/430m in browser with ORT Web / TF.js / tfjs-tflite?

    Hi, really exciting project! I'm wondering if you've published the model conversion script that you used to create the js_models files from the .pth model file? It would be awesome to see how the larger and newer models like RWKV-4 169m/430m perform in the browser! I think the inference speed of RWKV opens up many new possibilities for language models on the web.

    opened by josephrocca 32
  • CUDA compilation error with Ctx Length>2000

    CUDA compilation error with Ctx Length>2000

    Hello, I am trying out RWKV with audio modality and when I set T_MAX>>1000, it throws this error:

    Emitting ninja build file /root/.cache/torch_extensions/py39_cu116/timex/build.ninja...
Building extension module timex...
Allowing ninja to set a default number of workers... (overridable by setting the environment variable MAX_JOBS=N)
[1/2] /usr/local/cuda/bin/nvcc  -DTORCH_EXTENSION_NAME=timex -DTORCH_API_INCLUDE_EXTENSION_H -DPYBIND11_COMPILER_TYPE=\"_gcc\" -DPYBIND11_STDLIB=\"_libstdcpp\" -DPYBIND11_BUILD_ABI=\"_cxxabi1013\" -isystem /root/anaconda3/envs/surya-env/lib/python3.9/site-packages/torch/include -isystem /root/anaconda3/envs/surya-env/lib/python3.9/site-packages/torch/include/torch/csrc/api/include -isystem /root/anaconda3/envs/surya-env/lib/python3.9/site-packages/torch/include/TH -isystem /root/anaconda3/envs/surya-env/lib/python3.9/site-packages/torch/include/THC -isystem /usr/local/cuda/include -isystem /root/anaconda3/envs/surya-env/include/python3.9 -D_GLIBCXX_USE_CXX11_ABI=0 -D__CUDA_NO_HALF_OPERATORS__ -D__CUDA_NO_HALF_CONVERSIONS__ -D__CUDA_NO_BFLOAT16_CONVERSIONS__ -D__CUDA_NO_HALF2_OPERATORS__ --expt-relaxed-constexpr -gencode=arch=compute_80,code=compute_80 -gencode=arch=compute_80,code=sm_80 --compiler-options '-fPIC' --use_fast_math --extra-device-vectorization -DTmax=10000 -DBF=8 -DBB=2 -std=c++14 -c cuda/timex_cuda.cu -o timex_cuda.cuda.o 
FAILED: timex_cuda.cuda.o 
/usr/local/cuda/bin/nvcc  -DTORCH_EXTENSION_NAME=timex -DTORCH_API_INCLUDE_EXTENSION_H -DPYBIND11_COMPILER_TYPE=\"_gcc\" -DPYBIND11_STDLIB=\"_libstdcpp\" -DPYBIND11_BUILD_ABI=\"_cxxabi1013\" -isystem /root/anaconda3/envs/surya-env/lib/python3.9/site-packages/torch/include -isystem /root/anaconda3/envs/surya-env/lib/python3.9/site-packages/torch/include/torch/csrc/api/include -isystem /root/anaconda3/envs/surya-env/lib/python3.9/site-packages/torch/include/TH -isystem /root/anaconda3/envs/surya-env/lib/python3.9/site-packages/torch/include/THC -isystem /usr/local/cuda/include -isystem /root/anaconda3/envs/surya-env/include/python3.9 -D_GLIBCXX_USE_CXX11_ABI=0 -D__CUDA_NO_HALF_OPERATORS__ -D__CUDA_NO_HALF_CONVERSIONS__ -D__CUDA_NO_BFLOAT16_CONVERSIONS__ -D__CUDA_NO_HALF2_OPERATORS__ --expt-relaxed-constexpr -gencode=arch=compute_80,code=compute_80 -gencode=arch=compute_80,code=sm_80 --compiler-options '-fPIC' --use_fast_math --extra-device-vectorization -DTmax=10000 -DBF=8 -DBB=2 -std=c++14 -c cuda/timex_cuda.cu -o timex_cuda.cuda.o 
ptxas error   : Entry function '_Z15kernel_backwardIfEvPKT_S2_S2_PS0_S3_iii' uses too much shared data (0x30d40 bytes, 0xc000 max)
ptxas error   : Entry function '_Z14kernel_forwardIfEvPKT_S2_PS0_S0_iii' uses too much shared data (0x57e40 bytes, 0xc000 max)
ninja: build stopped: subcommand failed.

    GPU: A100, VRAM: 42GB, CUDA 11.6

    I am okay if the training takes a bit long. But I need this to work. Don't know any CUDA. Can you suggest some workarounds?

    Thanks for the incredible work btw!

    opened by ojus1 8
  • 关于调用模型做分类任务

    关于调用模型做分类任务

    你好作者！我对此工作很感兴趣，因为我现在在用基于transformer的模型做分类任务，transformer或者RNN在分类任务里通常采用最后一个模块的每个通道的最后一个元素作为输出，并通过全连接层映射到几个类别。 请问你觉得RWKV原理类似吗？依旧提取最后一个元素作为输出是否稳妥呢？希望您能给出一些建议，我将很感激！

    opened by louisinhit 2
Releases(4.00)
Owner
PENG Bo
http://zhihu.com/people/bopengbopeng
PENG Bo
GitHub Repository
Open Source Neural Machine Translation in PyTorch

OpenNMT-py: Open-Source Neural Machine Translation OpenNMT-py is the PyTorch version of the OpenNMT project, an open-source (MIT) neural machine trans

OpenNMT 5.8k Jan 04, 2023
precise iris segmentation

PI-DECODER Introduction PI-DECODER, a decoder structure designed for Precise Iris Segmentation and Location. The decoder structure is shown below: Ple

8 Aug 08, 2022
Hierarchical unsupervised and semi-supervised topic models for sparse count data with CorEx

Anchored CorEx: Hierarchical Topic Modeling with Minimal Domain Knowledge Correlation Explanation (CorEx) is a topic model that yields rich topics tha

Greg Ver Steeg 592 Dec 18, 2022
Open-source offline translation library written in Python. Uses OpenNMT for translations

Open source neural machine translation in Python. Designed to be used either as a Python library or desktop application. Uses OpenNMT for translations and PyQt for GUI.

Argos Open Tech 1.6k Jan 01, 2023
An open collection of annotated voices in Japanese language

声庭 (Koniwa): オープンな日本語音声とアノテーションのコレクション Koniwa (声庭): An open collection of annotated voices in Japanese language 概要 Koniwa(声庭)は利用・修正・再配布が自由でオープンな音声とアノテ

Koniwa project 32 Dec 14, 2022
Chinese named entity recognization (bert/roberta/macbert/bert_wwm with Keras)

Chinese named entity recognization (bert/roberta/macbert/bert_wwm with Keras)

2 Jul 05, 2022
This repository has a implementations of data augmentation for NLP for Japanese.

daaja This repository has a implementations of data augmentation for NLP for Japanese: EDA: Easy Data Augmentation Techniques for Boosting Performance

Koga Kobayashi 60 Nov 11, 2022
Code repository for "It's About Time: Analog clock Reading in the Wild"

it's about time Code repository for "It's About Time: Analog clock Reading in the Wild" Packages required: pytorch (used 1.9, any reasonable version s

52 Nov 10, 2022
A text file containing 479k English words for all your dictionary/word-based projects e.g: auto-completion / autosuggestion

List Of English Words A text file containing over 466k English words. While searching for a list of english words (for an auto-complete tutorial) I fo

dwyl 8.5k Jan 03, 2023
Snowball compiler and stemming algorithms

Snowball is a small string processing language for creating stemming algorithms for use in Information Retrieval, plus a collection of stemming algori

Snowball Stemming language and algorithms 613 Jan 07, 2023
Chinese NewsTitle Generation Project by GPT2.带有超级详细注释的中文GPT2新闻标题生成项目。

GPT2-NewsTitle 带有超详细注释的GPT2新闻标题生成项目 UpDate 01.02.2021 从网上收集数据，将清华新闻数据、搜狗新闻数据等新闻数据集，以及开源的一些摘要数据进行整理清洗，构建一个较完善的中文摘要数据集。 数据集清洗时，仅进行了简单地规则清洗。

logCong 785 Dec 29, 2022
Uncomplete archive of files from the European Nopsled Team

European Nopsled CTF Archive This is an archive of collected material from various Capture the Flag competitions that the European Nopsled team played

European Nopsled 4 Nov 24, 2021
Knowledge Management for Humans using Machine Learning & Tags

HyperTag helps humans intuitively express how they think about their files using tags and machine learning. Represent how you think using tags. Find what you look for using semantic search for your t

Ravn Tech, Inc. 166 Jan 07, 2023
Training RNNs as Fast as CNNs

News SRU++, a new SRU variant, is released. [tech report] [blog] The experimental code and SRU++ implementation are available on the dev branch which

Tao Lei 14 Dec 12, 2022
ANTLR (ANother Tool for Language Recognition) is a powerful parser generator for reading, processing, executing, or translating structured text or binary files.

ANTLR (ANother Tool for Language Recognition) is a powerful parser generator for reading, processing, executing, or translating structured text or binary files.

Antlr Project 13.6k Jan 05, 2023
Bnagla hand written document digiiztion

Bnagla hand written document digiiztion This repo addresses the problem of digiizing hand written documents in Bangla. Documents have definite fields

Mushfiqur Rahman 1 Dec 10, 2021
🤗 Transformers: State-of-the-art Natural Language Processing for Pytorch, TensorFlow, and JAX.

English | 简体中文 | 繁體中文 State-of-the-art Natural Language Processing for Jax, PyTorch and TensorFlow 🤗 Transformers provides thousands of pretrained mo

Hugging Face 77.2k Jan 03, 2023
multi-label，classifier，text classification，多标签文本分类，文本分类，BERT，ALBERT，multi-label-classification，seq2seq，attention，beam search

multi-label，classifier，text classification，多标签文本分类，文本分类，BERT，ALBERT，multi-label-classification，seq2seq，attention，beam search

hellonlp 30 Dec 12, 2022
Chatbot for the Chatango messaging platform

BroiestBot The baddest bot in the game right now. Uses the ch.py framework for joining Chantango rooms and responding to user messages. Commands If a

Todd Birchard 3 Jan 17, 2022
Code for the project carried out fulfilling the course requirements for Fall 2021 NLP at NYU

Introduction Fairseq(-py) is a sequence modeling toolkit that allows researchers and developers to train custom models for translation, summarization,

Sai Himal Allu 1 Apr 25, 2022