Rwkv 4 169m Pile

🚀 RWKV-4 | 基於Pile數據集訓練的1.69億參數模型卡片

RWKV是一個由Bo Peng領導的項目。你可以通過Johan Wind的博客文章這裡和這裡瞭解更多關於該模型架構的信息。你還可以通過加入RWKV Discord服務器來深入瞭解這個項目。

🚀 快速開始

模型簡述

以下是來自原倉庫的描述：

RWKV是一種具有Transformer級大語言模型性能的循環神經網絡（RNN）。它可以像GPT一樣直接進行訓練（可並行化）。它結合了RNN和Transformer的優點——性能出色、推理速度快、節省顯存、訓練速度快、具有“無限”上下文長度，並且能免費獲得句子嵌入。

✨ 主要特性

• 數據集：使用了EleutherAI/pile數據集進行訓練。

📚 詳細文檔

模型細節

模型架構的詳細信息可以在上述博客文章以及Hugging Face關於該模型集成的博客文章中找到。

模型使用

將原始權重轉換為Hugging Face格式

你可以使用convert_rwkv_checkpoint_to_hf.py腳本，通過指定原始權重的倉庫ID、文件名和輸出目錄來進行轉換。你還可以選擇通過傳遞--push_to_hub標誌和--model_name參數，將轉換後的模型直接推送到Hugging Face Hub上。

python convert_rwkv_checkpoint_to_hf.py --repo_id RAW_HUB_REPO --checkpoint_file RAW_FILE --output_dir OUTPUT_DIR --push_to_hub --model_name dummy_user/converted-rwkv

文本生成

你可以使用AutoModelForCausalLM和AutoTokenizer類從模型中生成文本。展開以下部分，瞭解如何在不同場景下運行該模型：

💻 使用示例

基礎用法

在CPU上運行模型

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("RWKV/rwkv-4-169m-pile")
tokenizer = AutoTokenizer.from_pretrained("RWKV/rwkv-4-169m-pile")

prompt = "\nIn a shocking finding, scientist discovered a herd of dragons living in a remote, previously unexplored valley, in Tibet. Even more surprising to the researchers was the fact that the dragons spoke perfect Chinese."

inputs = tokenizer(prompt, return_tensors="pt")
output = model.generate(inputs["input_ids"], max_new_tokens=40)
print(tokenizer.decode(output[0].tolist(), skip_special_tokens=True))

高級用法

在單個GPU上運行模型

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("RWKV/rwkv-4-169m-pile").to(0)
tokenizer = AutoTokenizer.from_pretrained("RWKV/rwkv-4-169m-pile")

prompt = "\nIn a shocking finding, scientist discovered a herd of dragons living in a remote, previously unexplored valley, in Tibet. Even more surprising to the researchers was the fact that the dragons spoke perfect Chinese."

inputs = tokenizer(prompt, return_tensors="pt").to(0)
output = model.generate(inputs["input_ids"], max_new_tokens=40)
print(tokenizer.decode(output[0].tolist(), skip_special_tokens=True))

在GPU上以半精度運行模型

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("RWKV/rwkv-4-169m-pile", torch_dtype=torch.float16).to(0)
tokenizer = AutoTokenizer.from_pretrained("RWKV/rwkv-4-169m-pile")

prompt = "\nIn a shocking finding, scientist discovered a herd of dragons living in a remote, previously unexplored valley, in Tibet. Even more surprising to the researchers was the fact that the dragons spoke perfect Chinese."

inputs = tokenizer(prompt, return_tensors="pt").to(0)
output = model.generate(inputs["input_ids"], max_new_tokens=40)
print(tokenizer.decode(output[0].tolist(), skip_special_tokens=True))

在多個GPU上運行模型

# pip install accelerate
from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("RWKV/rwkv-4-169m-pile", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained("RWKV/rwkv-4-169m-pile")

prompt = "\nIn a shocking finding, scientist discovered a herd of dragons living in a remote, previously unexplored valley, in Tibet. Even more surprising to the researchers was the fact that the dragons spoke perfect Chinese."

inputs = tokenizer(prompt, return_tensors="pt").to(0)
output = model.generate(inputs["input_ids"], max_new_tokens=40)
print(tokenizer.decode(output[0].tolist(), skip_special_tokens=True))

📄 許可證

如果你使用此模型，請考慮引用原倉庫此處的原始工作。