The RWKV Language Model

Last update: Jan 05, 2023

Related tags

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.

The Channel-mix is similar to GeGLU (https://arxiv.org/abs/2002.05202) with an extra R factor.
Finally, we add extra time-mixing as in (https://github.com/BlinkDL/minGPT-tuned). You can try reducing the amt of time-mixing in upper layers of deep models.

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:

(this is character-level loss with simplebooks-92 dataset https://dldata-public.s3.us-east-2.amazonaws.com/simplebooks.zip)

Comments

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

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?

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

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

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)

4.00(Dec 6, 2022)

Just a stable release.
Source code(tar.gz)
Source code(zip)
2.00(Mar 25, 2022)

Attached model : ctx1024-layer6-emb512 on enwik8 with 1.65 dev perplexity (0.72 BPC)
Source code(tar.gz)
Source code(zip)
enwik8-ppl1.65-6064-1024-RWKV-6-512-2022-03-25-21-05-13.zip(95.42 MB)
0.02(Aug 25, 2021)

v0.02 with RWKV, MHA_shift, MHA_rotary, MHA_pro, and time-shift mixing.
Source code(tar.gz)
Source code(zip)
0.01(Aug 13, 2021)

first release with RWKV, MHA_rotary, MHA_pro, and time-shift mixing.
Source code(tar.gz)
Source code(zip)

Owner

PENG Bo

http://zhihu.com/people/bopengbopeng

GitHub Repository

This repository contains Python scripts for extracting linguistic features from Filipino texts.

Filipino Text Linguistic Feature Extractors This repository contains scripts for extracting linguistic features from Filipino texts. The scripts were

1 Oct 05, 2021

State-of-the-art NLP through transformer models in a modular design and consistent APIs.

Trapper (Transformers wRAPPER) Trapper is an NLP library that aims to make it easier to train transformer based models on downstream tasks. It wraps h

42 Sep 21, 2022

MiCECo - Misskey Custom Emoji Counter

MiCECo Misskey Custom Emoji Counter Introduction This little script counts custo

7 Dec 25, 2022

Sentiment Classification using WSD, Maximum Entropy & Naive Bayes Classifiers

173 Jan 04, 2023

Turkish Stop Words Türkçe Dolgu Sözcükleri

trstop Turkish Stop Words Türkçe Dolgu Sözcükleri In this repository I put Turkish stop words that is contained in the first 10 thousand words with th

103 Nov 12, 2022

Code for Emergent Translation in Multi-Agent Communication

Emergent Translation in Multi-Agent Communication PyTorch implementation of the models described in the paper Emergent Translation in Multi-Agent Comm

75 Jul 15, 2022

Unet-TTS: Improving Unseen Speaker and Style Transfer in One-shot Voice Cloning

Unet-TTS: Improving Unseen Speaker and Style Transfer in One-shot Voice Cloning English | 中文 ❗ Now we provide inferencing code and pre-training models

164 Jan 02, 2023

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

1.5k Dec 05, 2022

Code for papers "Generation-Augmented Retrieval for Open-Domain Question Answering" and "Reader-Guided Passage Reranking for Open-Domain Question Answering", ACL 2021

This repo provides the code of the following papers: (GAR) "Generation-Augmented Retrieval for Open-domain Question Answering", ACL 2021 (RIDER) "Read

49 Dec 26, 2022

CATs: Semantic Correspondence with Transformers

CATs: Semantic Correspondence with Transformers For more information, check out the paper on [arXiv]. Training with different backbones and evaluation

74 Dec 10, 2021

A Fast Command Analyser based on Dict and Pydantic

Alconna Alconna 隶属于ArcletProject，在Cesloi内有内置 Alconna 是 Cesloi-CommandAnalysis 的高级版，支持解析消息链一般情况下请当作简易的消息链解析器/命令解析器文档暂时的文档 Example from arclet.alcon

19 Jan 03, 2023

Neural Lexicon Reader: Reduce Pronunciation Errors in End-to-end TTS by Leveraging External Textual Knowledge

Neural Lexicon Reader: Reduce Pronunciation Errors in End-to-end TTS by Leveraging External Textual Knowledge This is an implementation of the paper,

19 Oct 14, 2022

Code to reprudece NeurIPS paper: Accelerated Sparse Neural Training: A Provable and Efficient Method to Find N:M Transposable Masks

Accelerated Sparse Neural Training: A Provable and Efficient Method to FindN:M Transposable Masks Recently, researchers proposed pruning deep neural n

4 Feb 23, 2022

Code for our ACL 2021 (Findings) Paper - Fingerprinting Fine-tuned Language Models in the wild .

🌳 Fingerprinting Fine-tuned Language Models in the wild This is the code and dataset for our ACL 2021 (Findings) Paper - Fingerprinting Fine-tuned La

5 Sep 13, 2022

This repository contains the code, data, and models of the paper titled "CrossSum: Beyond English-Centric Cross-Lingual Abstractive Text Summarization for 1500+ Language Pairs".

CrossSum This repository contains the code, data, and models of the paper titled "CrossSum: Beyond English-Centric Cross-Lingual Abstractive Text Summ

29 Nov 19, 2022

sangha, pronounced "suhng-guh", is a social networking, booking platform where students and teachers can share their practice.

Flask React Project This is the backend for the Flask React project. Getting started Clone this repository (only this branch) git clone https://github

17 Sep 29, 2021

The RWKV Language Model

Related tags

Overview

RWKV-LM

Comments

Sequence to Sequence?

v4 model.py vs model_run.py

RWKV-4 169m/430m in browser with ORT Web / TF.js / tfjs-tflite?

CUDA compilation error with Ctx Length>2000

关于调用模型做分类任务