datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
lm-eval-data
LM Eval data
This repository, represents a giant collection of various lm-eval results built using the github scripts found here
The goal at the point of creation, was to collect as much LM-eval results for as many relevent model for the RWKV paper. For analysis.
Cloning the repo, without the jsonl dump
export GIT_LFS_SKIP_SMUDGE=1
git clone https://huggingface.co/datasets/rwkv-x-dev/lm-eval-data.git
Cloning the repo, with the archives (large)
# Clone with… See the full description on the dataset page: https://huggingface.co/datasets/rwkv-x-dev/lm-eval-data.rwkv-chatRWKV-Chat
RWKV-LM_pile_binidx_datasetRWKV-7-ArithmeticRWKV-7-Arithmetic-0.1B 加减法运算模型的训练和测试数据集。
该模型实现基础加减法运算和加减法方程求解功能，能够处理整数部分为 1-12 位、小数部分为 0-6 位的数值，支持中英文数字、全半角格式以及大小写字符的多种表示形式，可实现基础加减法运算和加减法方程求解功能。
训练数据集说明
以下是我们使用的加减法训练数据类型，共包含 30000587 33000147 条单轮加减法 QA 数据，约 1B（1014434168） token。
数据文件名
数据条数
数据说明
示例
ADD_4M
3997733
1. 使用‘全角’、‘中文数字’、‘大写中文数字’随机替换整个数字2. 运算符附近有 1~2 个随机空格3. 含简单自然语言描述/自然语言噪声
{"text": "User: 249476576 减 796580834 还剩多少？\n\nAssistant: -547104258"}
ADD_2M
1999673
1. 使用‘全角’、‘中文数字’、‘大写中文数字’随机替换整个数字2. 运算符附近有 1~2… See the full description on the dataset page: https://huggingface.co/datasets/shoumenchougou/RWKV-7-Arithmetic.blinkdl-rwkv-indonesiaEagleX-WorldContinued
Dataset Card for EagleX v2 Dataset
This dataset was used to train RWKV Eagle 7B for continued pretrain of 1.1T tokens (approximately) (boosting it to 2.25T) with the final model being released as RWKV EagleX v2.
Dataset Details
Dataset Description
EagleX-WorldContinued is a pretraining dataset built from many of our datasets over at Recursal AI + a few others.
Curated by: M8than, KaraKaraWitch, Darok
Funded by [optional]: Recursal.ai
Shared by [optional]:… See the full description on the dataset page: https://huggingface.co/datasets/RWKV/EagleX-WorldContinued.RWKV-World-v3
RWKV-7 (Goose) World v3 Corpus
Paper | Code
This is an itemised and annotated list of the RWKV World v3 corpus
which is a multilingual dataset with about 3.1T tokens used to train the
"Goose" RWKV-7 World model series.
RWKV World v3 was crafted from public datasets spanning >100 world languages
(80% English, 10% multilang, and 10% code). Also available as a HF Collection of Datasets.
Subsampled subsets (previews) of the corpus are available as 100k JSONL dataset and 1M JSONL dataset… See the full description on the dataset page: https://huggingface.co/datasets/Goose-World/RWKV-World-v3.details_KnutJaegersberg__RWKV-4-PilePlus-1B5-20230520-2942-486Gtokens-ctx4096
Dataset Card for Evaluation run of KnutJaegersberg/RWKV-4-PilePlus-1B5-20230520-2942-486Gtokens-ctx4096
Dataset Summary
Dataset automatically created during the evaluation run of model KnutJaegersberg/RWKV-4-PilePlus-1B5-20230520-2942-486Gtokens-ctx4096 on the Open LLM Leaderboard.
The dataset is composed of 3 configuration, each one coresponding to one of the evaluated task.
The dataset has been created from 2 run(s). Each run can be found as a specific split in each… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard-old/details_KnutJaegersberg__RWKV-4-PilePlus-1B5-20230520-2942-486Gtokens-ctx4096.RWKV-World-Listing
RWKV World Corpus
(includes v3, v2.1 and v2 subsets)
This is an itemised and annotated list of the RWKV World corpus as described in the RWKV-7 paper
which is a multilingual dataset with about 3.1T tokens used to train the
"Goose" RWKV-7 World model series.
RWKV World v3 was crafted from public datasets spanning >100 world languages
(80% English, 10% multilang, and 10% code).
PREVIEW
Random subsampled subsets of the world v3 corpus are available in the… See the full description on the dataset page: https://huggingface.co/datasets/RWKV/RWKV-World-Listing.RWKV-notebook-assets
RWKV notebook assets
Various asset files, used in RWKV notebook tutorial examples and demos
RWKV-TTS
Text-to-Speech Model based on RWKV7 Architecture
Introduction
This repository primarily explores the use of Focal Codec to convert between speech signals and tokens, and employs an RNN model based on RWKV7 to achieve token prediction for speech generation. The main features are as follows:
Multi-Stage Training: During the pre-training phase, the model performs Token Prediction merely on speech tokens, allowing it to extensively learn speech features from unlabeled speech… See the full description on the dataset page: https://huggingface.co/datasets/echodict/RWKV-TTS.openorca-gpt4OpenOrca but just the GPT4 bits.
rwkv-4-cpp-quantize-binYou are probably looking for the raven models found here : https://huggingface.co/BlinkDL/rwkv-4-raven
This is a collection of converted CPP binaries, that maybe prequantized.
This is primarily used for the rwkv-cpp-node project here : https://github.com/RWKV/RWKV-cpp-node
details_KnutJaegersberg__RWKV-4-PilePlus-169M-20230520-done-ctx4096
Dataset Card for Evaluation run of KnutJaegersberg/RWKV-4-PilePlus-169M-20230520-done-ctx4096
Dataset Summary
Dataset automatically created during the evaluation run of model KnutJaegersberg/RWKV-4-PilePlus-169M-20230520-done-ctx4096 on the Open LLM Leaderboard.
The dataset is composed of 64 configuration, each one coresponding to one of the evaluated task.
The dataset has been created from 2 run(s). Each run can be found as a specific split in each configuration, the… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard-old/details_KnutJaegersberg__RWKV-4-PilePlus-169M-20230520-done-ctx4096.rwkv-world-3-subsample-previewlaces-owt-rwkv-tokens
LACES training data: OpenWebText RWKV tokens (512)
Preprocessed OpenWebText token shards for training LACES (Linear-Attention Continuous-state DiffuSion) — the DiffRWKV State-Hijacking RELAY 2.9B 512-token trajectory champion recipe.
304,052 samples, max length 512, tokenized with the RWKV7-Goose-World3 tokenizer
Format: one *_tokens.npz per sample
Usage
hf download humanlong/laces-owt-rwkv-tokens --repo-type dataset --local-dir .
tar xzf… See the full description on the dataset page: https://huggingface.co/datasets/humanlong/laces-owt-rwkv-tokens.details_RWKV__rwkv-4-14b-pile
Dataset Card for Evaluation run of RWKV/rwkv-4-14b-pile
Dataset Summary
Dataset automatically created during the evaluation run of model RWKV/rwkv-4-14b-pile on the Open LLM Leaderboard.
The dataset is composed of 64 configuration, each one coresponding to one of the evaluated task.
The dataset has been created from 2 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard-old/details_RWKV__rwkv-4-14b-pile.details_RWKV__rwkv-raven-3b
Dataset Card for Evaluation run of RWKV/rwkv-raven-3b
Dataset Summary
Dataset automatically created during the evaluation run of model RWKV/rwkv-raven-3b on the Open LLM Leaderboard.
The dataset is composed of 64 configuration, each one coresponding to one of the evaluated task.
The dataset has been created from 3 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard-old/details_RWKV__rwkv-raven-3b.rwkv_synthetic_project_unofficial
RWKV Synthetic Project (Unofficial)
An unofficial community-driven collection of niche knowledge, experiments, discussions, and notes related to RWKV.
This dataset is intended to preserve and organize interesting RWKV-related material that may be too specific, experimental, or informal to fit elsewhere.
What can be contributed?
If you have something interesting about RWKV, feel free to share it in the Discussions.
For example:
Niche RWKV-related questions and… See the full description on the dataset page: https://huggingface.co/datasets/Ilikemechuri/rwkv_synthetic_project_unofficial.mod-rwkv-instruction-oldmod-rwkv-instruct-oigmoderationdetails_RWKV__rwkv-raven-7b
Dataset Card for Evaluation run of RWKV/rwkv-raven-7b
Dataset Summary
Dataset automatically created during the evaluation run of model RWKV/rwkv-raven-7b on the Open LLM Leaderboard.
The dataset is composed of 64 configuration, each one coresponding to one of the evaluated task.
The dataset has been created from 2 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard-old/details_RWKV__rwkv-raven-7b.details_KnutJaegersberg__RWKV-4-PilePlus-430M-20230520-6162-1018Gtokens-ctx4098
Dataset Card for Evaluation run of KnutJaegersberg/RWKV-4-PilePlus-430M-20230520-6162-1018Gtokens-ctx4098
Dataset Summary
Dataset automatically created during the evaluation run of model KnutJaegersberg/RWKV-4-PilePlus-430M-20230520-6162-1018Gtokens-ctx4098 on the Open LLM Leaderboard.
The dataset is composed of 64 configuration, each one coresponding to one of the evaluated task.
The dataset has been created from 2 run(s). Each run can be found as a specific split in… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard-old/details_KnutJaegersberg__RWKV-4-PilePlus-430M-20230520-6162-1018Gtokens-ctx4098.details_RWKV__rwkv-4-1b5-pile
Dataset Card for Evaluation run of RWKV/rwkv-4-1b5-pile
Dataset Summary
Dataset automatically created during the evaluation run of model RWKV/rwkv-4-1b5-pile on the Open LLM Leaderboard.
The dataset is composed of 64 configuration, each one coresponding to one of the evaluated task.
The dataset has been created from 2 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard-old/details_RWKV__rwkv-4-1b5-pile.slimpajama-binidxdetails_RWKV__rwkv-4-430m-pile
Dataset Card for Evaluation run of RWKV/rwkv-4-430m-pile
Dataset Summary
Dataset automatically created during the evaluation run of model RWKV/rwkv-4-430m-pile on the Open LLM Leaderboard.
The dataset is composed of 64 configuration, each one coresponding to one of the evaluated task.
The dataset has been created from 2 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard-old/details_RWKV__rwkv-4-430m-pile.details_RWKV__rwkv-4-169m-pile
Dataset Card for Evaluation run of RWKV/rwkv-4-169m-pile
Dataset Summary
Dataset automatically created during the evaluation run of model RWKV/rwkv-4-169m-pile on the Open LLM Leaderboard.
The dataset is composed of 64 configuration, each one coresponding to one of the evaluated task.
The dataset has been created from 2 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard-old/details_RWKV__rwkv-4-169m-pile.details_RWKV__rwkv-4-3b-pile
Dataset Card for Evaluation run of RWKV/rwkv-4-3b-pile
Dataset Summary
Dataset automatically created during the evaluation run of model RWKV/rwkv-4-3b-pile on the Open LLM Leaderboard.
The dataset is composed of 64 configuration, each one coresponding to one of the evaluated task.
The dataset has been created from 2 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard-old/details_RWKV__rwkv-4-3b-pile.rwkv-600m-sft-base-clean-data-v7-base-clean