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
The PyTorch based implementation of continuous integrate-and-fire (CIF) module.

CIF-PyTorch This is a PyTorch based implementation of continuous integrate-and-fire (CIF) module for end-to-end (E2E) automatic speech recognition (AS

Minglun Han 24 Dec 29, 2022
Bot to connect a real Telegram user, simulating responses with OpenAI's davinci GPT-3 model.

AI-BOT Bot to connect a real Telegram user, simulating responses with OpenAI's davinci GPT-3 model.

Thempra 2 Dec 21, 2022
STT for TorchScript is a port of Coqui STT based on DeepSpeech to PyTorch.

st3 STT for TorchScript is a port of Coqui STT based on DeepSpeech to PyTorch. Currently it supports converting pbmm models to pt scripts with integra

Vlad Ki 8 Oct 18, 2021
Lattice methods in TensorFlow

TensorFlow Lattice TensorFlow Lattice is a library that implements constrained and interpretable lattice based models. It is an implementation of Mono

504 Dec 20, 2022
Code for the ACL 2021 paper "Structural Guidance for Transformer Language Models"

Structural Guidance for Transformer Language Models This repository accompanies the paper, Structural Guidance for Transformer Language Models, publis

International Business Machines 10 Dec 14, 2022
Unofficial Parallel WaveGAN (+ MelGAN & Multi-band MelGAN & HiFi-GAN & StyleMelGAN) with Pytorch

Parallel WaveGAN implementation with Pytorch This repository provides UNOFFICIAL pytorch implementations of the following models: Parallel WaveGAN Mel

Tomoki Hayashi 1.2k Dec 23, 2022
Count the frequency of letters or words in a text file and show a graph.

Word Counter By EBUS Coding Club Count the frequency of letters or words in a text file and show a graph. Requirements Python 3.9 or higher matplotlib

EBUS Coding Club 0 Apr 09, 2022
Framework for fine-tuning pretrained transformers for Named-Entity Recognition (NER) tasks

NERDA Not only is NERDA a mesmerizing muppet-like character. NERDA is also a python package, that offers a slick easy-to-use interface for fine-tuning

Ekstra Bladet 141 Dec 30, 2022
The training code for the 4th place model at MDX 2021 leaderboard A.

The training code for the 4th place model at MDX 2021 leaderboard A.

Chin-Yun Yu 32 Dec 18, 2022
topic modeling on unstructured data in Space news articles retrieved from the Guardian (UK) newspaper using API

NLP Space News Topic Modeling Photos by nasa.gov (1, 2, 3, 4, 5) and extremetech.com Table of Contents Project Idea Data acquisition Primary data sour

edesz 1 Jan 03, 2022
CLIPfa: Connecting Farsi Text and Images

CLIPfa: Connecting Farsi Text and Images OpenAI released the paper Learning Transferable Visual Models From Natural Language Supervision in which they

Sajjad Ayoubi 66 Dec 14, 2022
PyTorch implementation of the paper: Text is no more Enough! A Benchmark for Profile-based Spoken Language Understanding

Text is no more Enough! A Benchmark for Profile-based Spoken Language Understanding This repository contains the official PyTorch implementation of th

Xiao Xu 26 Dec 14, 2022
Implementing SimCSE(paper, official repository) using TensorFlow 2 and KR-BERT.

KR-BERT-SimCSE Implementing SimCSE(paper, official repository) using TensorFlow 2 and KR-BERT. Training Unsupervised python train_unsupervised.py --mi

Jeong Ukjae 27 Dec 12, 2022
This repository will contain the code for the CVPR 2021 paper "GIRAFFE: Representing Scenes as Compositional Generative Neural Feature Fields"

GIRAFFE: Representing Scenes as Compositional Generative Neural Feature Fields Project Page | Paper | Supplementary | Video | Slides | Blog | Talk If

1.1k Dec 27, 2022
Web Scraping, Document Deduplication & GPT-2 Fine-tuning with a newly created scam dataset.

Web Scraping, Document Deduplication & GPT-2 Fine-tuning with a newly created scam dataset.

18 Nov 28, 2022
Officile code repository for "A Game-Theoretic Perspective on Risk-Sensitive Reinforcement Learning"

CvarAdversarialRL Official code repository for "A Game-Theoretic Perspective on Risk-Sensitive Reinforcement Learning". Initial setup Create a virtual

Mathieu Godbout 1 Nov 19, 2021
Code for "Semantic Role Labeling as Dependency Parsing: Exploring Latent Tree Structures Inside Arguments".

Code for "Semantic Role Labeling as Dependency Parsing: Exploring Latent Tree Structures Inside Arguments".

Yu Zhang 50 Nov 08, 2022
Named Entity Recognition API used by TEI Publisher

TEI Publisher Named Entity Recognition API This repository contains the API used by TEI Publisher's web-annotation editor to detect entities in the in

e-editiones.org 14 Nov 15, 2022
This is a MD5 password/passphrase brute force tool

CROWES-PASS-CRACK-TOOl This is a MD5 password/passphrase brute force tool How to install: Do 'git clone https://github.com/CROW31/CROWES-PASS-CRACK-TO

9 Mar 02, 2022
Train and use generative text models in a few lines of code.

blather Train and use generative text models in a few lines of code. To see blather in action check out the colab notebook! Installation Use the packa

Dan Carroll 16 Nov 07, 2022