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
Official implementation of Meta-StyleSpeech and StyleSpeech

Meta-StyleSpeech : Multi-Speaker Adaptive Text-to-Speech Generation Dongchan Min, Dong Bok Lee, Eunho Yang, and Sung Ju Hwang This is an official code

min95 169 Jan 05, 2023
HAN2HAN : Hangul Font Generation

HAN2HAN : Hangul Font Generation

Changwoo Lee 36 Dec 28, 2022
Mkdocs + material + cool stuff

Modern-Python-Doc-Example mkdocs + material + cool stuff Doc is live here Features out of the box amazing good looking website thanks to mkdocs.org an

Francesco Saverio Zuppichini 61 Oct 26, 2022
Fidibo.com comments Sentiment Analyser

Fidibo.com comments Sentiment Analyser Introduction This project first asynchronously grab Fidibo.com books comment data using grabber.py and then sav

Iman Kermani 3 Apr 15, 2022
Curso práctico: NLP de cero a cien 🤗

Curso Práctico: NLP de cero a cien Comprende todos los conceptos y arquitecturas clave del estado del arte del NLP y aplícalos a casos prácticos utili

Somos NLP 147 Jan 06, 2023
Study German declensions (dER nettE Mann, ein nettER Mann, mit dEM nettEN Mann, ohne dEN nettEN Mann ...) Generate as many exercises as you want using the incredible power of SPACY!

Study German declensions (dER nettE Mann, ein nettER Mann, mit dEM nettEN Mann, ohne dEN nettEN Mann ...) Generate as many exercises as you want using the incredible power of SPACY!

Hans Alemão 4 Jul 20, 2022
Fine-tuning scripts for evaluating transformer-based models on KLEJ benchmark.

The KLEJ Benchmark Baselines The KLEJ benchmark (Kompleksowa Lista Ewaluacji Językowych) is a set of nine evaluation tasks for the Polish language und

Allegro Tech 17 Oct 18, 2022
Kurumi ChatBot

KurumiChatBot Just another Telegram AI chat bot written in Python using Pyrogram. A public running instance can be found on telegram as @TokisakiChatB

Yoga Pranata 3 Jun 28, 2022
Smart discord chatbot integrated with Dialogflow to manage different classrooms and assist in teaching!

smart-school-chatbot Smart discord chatbot integrated with Dialogflow to interact with students naturally and manage different classes in a school. De

Tom Huynh 5 Oct 24, 2022
Colibri core is an NLP tool as well as a C++ and Python library for working with basic linguistic constructions such as n-grams and skipgrams (i.e patterns with one or more gaps, either of fixed or dynamic size) in a quick and memory-efficient way. At the core is the tool ``colibri-patternmodeller`` whi ch allows you to build, view, manipulate and query pattern models.

Colibri Core by Maarten van Gompel, [email protected], Radboud University Nij

Maarten van Gompel 122 Nov 17, 2022
Multilingual finetuning of Machine Translation model on low-resource languages. Project for Deep Natural Language Processing course.

Low-resource-Machine-Translation This repository contains the code for the project relative to the course Deep Natural Language Processing. The goal o

Andrea Cavallo 3 Jun 22, 2022
Transformer related optimization, including BERT, GPT

This repository provides a script and recipe to run the highly optimized transformer-based encoder and decoder component, and it is tested and maintained by NVIDIA.

NVIDIA Corporation 1.7k Jan 04, 2023
Concept Modeling: Topic Modeling on Images and Text

Concept is a technique that leverages CLIP and BERTopic-based techniques to perform Concept Modeling on images.

Maarten Grootendorst 120 Dec 27, 2022
fastai ulmfit - Pretraining the Language Model, Fine-Tuning and training a Classifier

fast.ai ULMFiT with SentencePiece from pretraining to deployment Motivation: Why even bother with a non-BERT / Transformer language model? Short answe

Florian Leuerer 26 May 27, 2022
OCR을 이용하여 인원수를 인식 후 줌을 Kill 해줍니다

How To Use killtheZoom-2.0 Windows 0. https://joyhong.tistory.com/79 이 글을 보면서 tesseract를 C:\Program Files\Tesseract-OCR 경로로 설치해주세요(한국어 언어 추가 필요) 상단의 초

김정인 9 Sep 13, 2021
Signature remover is a NLP based solution which removes email signatures from the rest of the text.

Signature Remover Signature remover is a NLP based solution which removes email signatures from the rest of the text. It helps to enchance data conten

Forges Alterway 8 Jan 06, 2023
Unsupervised intent recognition

INTENT author: steeve LAQUITAINE description: deployment pattern: currently batch only Setup & run git clone https://github.com/slq0/intent.git bash

sl 1 Apr 08, 2022
Include MelGAN, HifiGAN and Multiband-HifiGAN, maybe NHV in the future.

Fast (GAN Based Neural) Vocoder Chinese README Todo Submit demo Support NHV Discription Include MelGAN, HifiGAN and Multiband-HifiGAN, maybe include N

Zhengxi Liu (刘正曦) 134 Dec 16, 2022
Multilingual text (NLP) processing toolkit

polyglot Polyglot is a natural language pipeline that supports massive multilingual applications. Free software: GPLv3 license Documentation: http://p

RAMI ALRFOU 2.1k Jan 07, 2023
A natural language processing model for sequential sentence classification in medical abstracts.

NLP PubMed Medical Research Paper Abstract (Randomized Controlled Trial) A natural language processing model for sequential sentence classification in

Hemanth Chandran 1 Jan 17, 2022