The RWKV Language Model

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
DLO8012: Natural Language Processing & CSL804: Computational Lab - II

NATURAL-LANGUAGE-PROCESSING-AND-COMPUTATIONAL-LAB-II DLO8012: NLP & CSL804: CL-II [SEMESTER VIII] Syllabus NLP - Reference Books THE WALL MEGA SATISH

AMEY THAKUR 7 Apr 28, 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
Train 🤗transformers with DeepSpeed: ZeRO-2, ZeRO-3

Fork from https://github.com/huggingface/transformers/tree/86d5fb0b360e68de46d40265e7c707fe68c8015b/examples/pytorch/language-modeling at 2021.05.17.

Junbum Lee 12 Oct 26, 2022
Translate U is capable of translating the text present in an image from one language to the other.

Translate U is capable of translating the text present in an image from one language to the other. The app uses OCR and Google translate to identify and translate across 80+ languages.

Neelanjan Manna 1 Dec 22, 2021
Natural language Understanding Toolkit

Natural language Understanding Toolkit TOC Requirements Installation Documentation CLSCL NER References Requirements To install nut you need: Python 2

Peter Prettenhofer 119 Oct 08, 2022
PyTorch implementation of Tacotron speech synthesis model.

tacotron_pytorch PyTorch implementation of Tacotron speech synthesis model. Inspired from keithito/tacotron. Currently not as much good speech quality

Ryuichi Yamamoto 279 Dec 09, 2022
The model is designed to train a single and large neural network in order to predict correct translation by reading the given sentence.

Neural Machine Translation communication system The model is basically direct to convert one source language to another targeted language using encode

Nishant Banjade 7 Sep 22, 2022
Code for our ACL 2021 paper - ConSERT: A Contrastive Framework for Self-Supervised Sentence Representation Transfer

ConSERT Code for our ACL 2021 paper - ConSERT: A Contrastive Framework for Self-Supervised Sentence Representation Transfer Requirements torch==1.6.0

Yan Yuanmeng 478 Dec 25, 2022
Python functions for summarizing and improving voice dictation input.

Helpmespeak Help me speak uses Python functions for summarizing and improving voice dictation input. Get started with OpenAI gpt-3 OpenAI is a amazing

Margarita Humanitarian Foundation 6 Dec 17, 2022
ConvBERT-Prod

ConvBERT 目录 0. 仓库结构 1. 简介 2. 数据集和复现精度 3. 准备数据与环境 3.1 准备环境 3.2 准备数据 3.3 准备模型 4. 开始使用 4.1 模型训练 4.2 模型评估 4.3 模型预测 5. 模型推理部署 5.1 基于Inference的推理 5.2 基于Serv

yujun 7 Apr 08, 2022
MRC approach for Aspect-based Sentiment Analysis (ABSA)

B-MRC MRC approach for Aspect-based Sentiment Analysis (ABSA) Paper: Bidirectional Machine Reading Comprehension for Aspect Sentiment Triplet Extracti

Phuc Phan 1 Apr 05, 2022
File-based TF-IDF: Calculates keywords in a document, using a word corpus.

File-based TF-IDF Calculates keywords in a document, using a word corpus. Why? Because I found myself with hundreds of plain text files, with no way t

Jakob Lindskog 1 Feb 11, 2022
Automatically search Stack Overflow for the command you want to run

stackshell Automatically search Stack Overflow (and other Stack Exchange sites) for the command you want to ru Use the up and down arrows to change be

circuit10 22 Oct 27, 2021
Bu Chatbot, Konya Bilim Merkezi Yen için tasarlanmış olan bir projedir.

chatbot Bu Chatbot, Konya Bilim Merkezi Yeni Ufuklar Sergisi için 2021 Yılında tasarlanmış olan bir projedir. Chatbot Python ortamında yazılmıştır. Sö

Emre Özkul 1 Feb 23, 2022
ELECTRA: Pre-training Text Encoders as Discriminators Rather Than Generators

ELECTRA Introduction ELECTRA is a method for self-supervised language representation learning. It can be used to pre-train transformer networks using

Google Research 2.1k Dec 28, 2022
Official implementation of MLP Singer: Towards Rapid Parallel Korean Singing Voice Synthesis

MLP Singer Official implementation of MLP Singer: Towards Rapid Parallel Korean Singing Voice Synthesis. Audio samples are available on our demo page.

Neosapience 103 Dec 23, 2022
结巴中文分词

jieba “结巴”中文分词:做最好的 Python 中文分词组件 "Jieba" (Chinese for "to stutter") Chinese text segmentation: built to be the best Python Chinese word segmentation

Sun Junyi 29.8k Jan 02, 2023
Download videos from YouTube/Twitch/Twitter right in the Windows Explorer, without installing any shady shareware apps

youtube-dl and ffmpeg Windows Explorer Integration Download videos from YouTube/Twitch/Twitter and more (any platform that is supported by youtube-dl)

Wolfgang 226 Dec 30, 2022
The entmax mapping and its loss, a family of sparse softmax alternatives.

entmax This package provides a pytorch implementation of entmax and entmax losses: a sparse family of probability mappings and corresponding loss func

DeepSPIN 330 Dec 22, 2022
An easy-to-use framework for BERT models, with trainers, various NLP tasks and detailed annonations

FantasyBert English | 中文 Introduction An easy-to-use framework for BERT models, with trainers, various NLP tasks and detailed annonations. You can imp

Fan 137 Oct 26, 2022