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
TunBERT is the first release of a pre-trained BERT model for the Tunisian dialect using a Tunisian Common-Crawl-based dataset.

TunBERT is the first release of a pre-trained BERT model for the Tunisian dialect using a Tunisian Common-Crawl-based dataset. TunBERT was applied to three NLP downstream tasks: Sentiment Analysis (S

InstaDeep Ltd 72 Dec 09, 2022
Longformer: The Long-Document Transformer

Longformer Longformer and LongformerEncoderDecoder (LED) are pretrained transformer models for long documents. ***** New December 1st, 2020: Longforme

AI2 1.6k Dec 29, 2022
Common Voice Dataset explorer

Common Voice Dataset Explorer Common Voice Dataset is by Mozilla Made during huggingface finetuning week Usage pip install -r requirements.txt streaml

Ceyda Cinarel 22 Nov 16, 2022
NLPretext packages in a unique library all the text preprocessing functions you need to ease your NLP project.

NLPretext packages in a unique library all the text preprocessing functions you need to ease your NLP project.

Artefact 114 Dec 15, 2022
Toward Model Interpretability in Medical NLP

Toward Model Interpretability in Medical NLP LING380: Topics in Computational Linguistics Final Project James Cross ( 1 Mar 04, 2022

This is a NLP based project to extract effective date of the contract from their text files.

Date-Extraction-from-Contracts This is a NLP based project to extract effective date of the contract from their text files. Problem statement This is

Sambhav Garg 1 Jan 26, 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
BERT score for text generation

BERTScore Automatic Evaluation Metric described in the paper BERTScore: Evaluating Text Generation with BERT (ICLR 2020). News: Features to appear in

Tianyi 1k Jan 08, 2023
Modeling cumulative cases of Covid-19 in the US during the Covid 19 Delta wave using Bayesian methods.

Introduction The goal of this analysis is to find a model that fits the observed cumulative cases of COVID-19 in the US, starting in Mid-July 2021 and

Alexander Keeney 1 Jan 05, 2022
Ongoing research training transformer language models at scale, including: BERT & GPT-2

What is this fork of Megatron-LM and Megatron-DeepSpeed This is a detached fork of https://github.com/microsoft/Megatron-DeepSpeed, which in itself is

BigScience Workshop 316 Jan 03, 2023
मराठी भाषा वाचविण्याचा एक प्रयास. इंग्रजी ते मराठीचा शब्दकोश. An attempt to preserve the Marathi language. A lightweight and ad free English to Marathi thesaurus.

For English, scroll down मराठी शब्द मराठी भाषा वाचवण्यासाठी मी हा ओपन सोर्स प्रोजेक्ट सुरू केला आहे. माझ्या मते, आपली भाषा हळूहळू आणि कोणाचाही लक्षात

मुक्त स्त्रोत 20 Oct 11, 2022
A Persian Image Captioning model based on Vision Encoder Decoder Models of the transformers🤗.

Persian-Image-Captioning We fine-tuning the Vision Encoder Decoder Model for the task of image captioning on the coco-flickr-farsi dataset. The implem

Hamtech-ai 15 Aug 25, 2022
Transfer Learning from Speaker Verification to Multispeaker Text-To-Speech Synthesis (SV2TTS)

This repository is an implementation of Transfer Learning from Speaker Verification to Multispeaker Text-To-Speech Synthesis (SV2TTS) with a vocoder that works in real-time. Feel free to check my the

Corentin Jemine 38.5k Jan 03, 2023
Twitter-Sentiment-Analysis - Twitter sentiment analysis for india's top online retailers(2019 to 2022)

Twitter-Sentiment-Analysis Twitter sentiment analysis for india's top online retailers(2019 to 2022) Project Overview : Sentiment Analysis helps us to

Balaji R 1 Jan 01, 2022
An open-source NLP research library, built on PyTorch.

An Apache 2.0 NLP research library, built on PyTorch, for developing state-of-the-art deep learning models on a wide variety of linguistic tasks. Quic

AI2 11.4k Jan 01, 2023
Code for PED: DETR For (Crowd) Pedestrian Detection

Code for PED: DETR For (Crowd) Pedestrian Detection

36 Sep 13, 2022
Simple Python library, distributed via binary wheels with few direct dependencies, for easily using wav2vec 2.0 models for speech recognition

Wav2Vec2 STT Python Beta Software Simple Python library, distributed via binary wheels with few direct dependencies, for easily using wav2vec 2.0 mode

David Zurow 22 Dec 29, 2022
Bidirectional LSTM-CRF and ELMo for Named-Entity Recognition, Part-of-Speech Tagging and so on.

anaGo anaGo is a Python library for sequence labeling(NER, PoS Tagging,...), implemented in Keras. anaGo can solve sequence labeling tasks such as nam

Hiroki Nakayama 1.5k Dec 05, 2022
This repository structures data in title, summary, tags, sentiment given a fragment of a conversation

Understand-conversation-AI This repository structures data in title, summary, tags, sentiment given a fragment of a conversation How to install: pip i

Juan Camilo López Montes 1 Jan 11, 2022
KoBERTopic은 BERTopic을 한국어 데이터에 적용할 수 있도록 토크나이저와 BERT를 수정한 코드입니다.

KoBERTopic 모델 소개 KoBERTopic은 BERTopic을 한국어 데이터에 적용할 수 있도록 토크나이저와 BERT를 수정했습니다. 기존 BERTopic : https://github.com/MaartenGr/BERTopic/tree/05a6790b21009d

Won Joon Yoo 26 Jan 03, 2023