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
This library is testing the ethics of language models by using natural adversarial texts.

prompt2slip This library is testing the ethics of language models by using natural adversarial texts. This tool allows for short and simple code and v

9 Dec 28, 2021
Python Implementation of ``Modeling the Influence of Verb Aspect on the Activation of Typical Event Locations with BERT'' (Findings of ACL: ACL 2021)

BERT-for-Surprisal Python Implementation of ``Modeling the Influence of Verb Aspect on the Activation of Typical Event Locations with BERT'' (Findings

7 Dec 05, 2022
Collection of scripts to pinpoint obfuscated code

Obfuscation Detection (v1.0) Author: Tim Blazytko Automatically detect control-flow flattening and other state machines Description: Scripts and binar

Tim Blazytko 230 Nov 26, 2022
A spaCy wrapper of OpenTapioca for named entity linking on Wikidata

spaCyOpenTapioca A spaCy wrapper of OpenTapioca for named entity linking on Wikidata. Table of contents Installation How to use Local OpenTapioca Vizu

Universitätsbibliothek Mannheim 80 Jan 03, 2023
✔👉A Centralized WebApp to Ensure Road Safety by checking on with the activities of the driver and activating label generator using NLP.

AI-For-Road-Safety Challenge hosted by Omdena Hyderabad Chapter Original Repo Link : https://github.com/OmdenaAI/omdena-india-roadsafety Final Present

Prathima Kadari 7 Nov 29, 2022
Official PyTorch implementation of "Dual Path Learning for Domain Adaptation of Semantic Segmentation".

Dual Path Learning for Domain Adaptation of Semantic Segmentation Official PyTorch implementation of "Dual Path Learning for Domain Adaptation of Sema

27 Dec 22, 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
基于pytorch+bert的中文事件抽取

pytorch_bert_event_extraction 基于pytorch+bert的中文事件抽取,主要思想是QA(问答)。 要预先下载好chinese-roberta-wwm-ext模型,并在运行时指定模型的位置。

西西嘛呦 31 Nov 30, 2022
AI-Broad-casting - AI Broad casting with python

Basic Code 1. Use The Code Configuration Environment conda create -n code_base p

Trains an OpenNMT PyTorch model and SentencePiece tokenizer.

Trains an OpenNMT PyTorch model and SentencePiece tokenizer. Designed for use with Argos Translate and LibreTranslate.

Argos Open Tech 61 Dec 13, 2022
Unsupervised text tokenizer for Neural Network-based text generation.

SentencePiece SentencePiece is an unsupervised text tokenizer and detokenizer mainly for Neural Network-based text generation systems where the vocabu

Google 6.4k Jan 01, 2023
Document processing using transformers

Doc Transformers Document processing using transformers. This is still in developmental phase, currently supports only extraction of form data i.e (ke

Vishnu Nandakumar 13 Dec 21, 2022
Smart discord chatbot integrated with Dialogflow

academic-NLP-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
Natural language computational chemistry command line interface.

nlcc Install pip install nlcc Must have Open-AI Codex key: export OPENAI_API_KEY=your key here then nlcc key bindings ctrl-w copy to clipboard (Note

Andrew White 37 Dec 14, 2022
GraphNLI: A Graph-based Natural Language Inference Model for Polarity Prediction in Online Debates

GraphNLI: A Graph-based Natural Language Inference Model for Polarity Prediction in Online Debates Vibhor Agarwal, Sagar Joglekar, Anthony P. Young an

Vibhor Agarwal 2 Jun 30, 2022
Blackstone is a spaCy model and library for processing long-form, unstructured legal text

Blackstone Blackstone is a spaCy model and library for processing long-form, unstructured legal text. Blackstone is an experimental research project f

ICLR&D 579 Jan 08, 2023
🤗 The largest hub of ready-to-use NLP datasets for ML models with fast, easy-to-use and efficient data manipulation tools

🤗 The largest hub of ready-to-use NLP datasets for ML models with fast, easy-to-use and efficient data manipulation tools

Hugging Face 15k Jan 02, 2023
Parrot is a paraphrase based utterance augmentation framework purpose built to accelerate training NLU models

Parrot is a paraphrase based utterance augmentation framework purpose built to accelerate training NLU models. A paraphrase framework is more than just a paraphrasing model.

Prithivida 681 Jan 01, 2023
Shirt Bot is a discord bot which uses GPT-3 to generate text

SHIRT BOT · Shirt Bot is a discord bot which uses GPT-3 to generate text. Made by Cyclcrclicly#3420 (474183744685604865) on Discord. Support Server EX

31 Oct 31, 2022
AMUSE - financial summarization

AMUSE AMUSE - financial summarization Unzip data.zip Train new model: python FinAnalyze.py --task train --start 0 --count how many files,-1 for all

1 Jan 11, 2022