dean2023/cn-roberta-wwm-ext-car
By dean2023 · ↓ 0 · ♥ 0 · Updated 2025-02-23
What's in this repo
This page lists the files in dean2023/cn-roberta-wwm-ext-car, including cn_roberta_v1.onnx.
- 1 ONNX models (.onnx files)
- 8 files, including 3 .json, 1 .md, 1 .onnx, 1 .rknn, 1 .txt
- Last updated 2025-02-23
ONNX files
| File | Size | Download |
|---|---|---|
| cn_roberta_v1.onnx | — | Download |
How to use
These ONNX files can be loaded with ONNX Runtime. To inspect a model graph, upload the file to the Netron viewer. Download links point to the original files on Hugging Face.
README
Chinese-RoBERTa-wwm-ext RKNN Deployment Model
基于 hfl/chinese-roberta-wwm-ext 的领域适配训练模型，支持ONNX推理和RK3588芯片部署。
模型概述
本模型是基于 RoBERTa-wwm-ext 架构，在汽车领域中文文本数据上进行微调的预训练语言模型，特点包括：
- 支持动态输入长度（最大 128 tokens）
- 提供PyTorch/ONNX/RKNN三种格式
- 针对RK3588芯片进行内核优化
主要特性
✅ 领域适配：使用汽车论坛、维修手册等专业语料微调
✅ 部署友好：提供量化后的ONNX/RKNN模型
✅ 硬件加速：RKNN模型针对NPU计算单元优化
硬件要求
| 格式 | 最低要求 |
|---|---|
| PyTorch | CPU / NVIDIA GPU (显存 ≥4GB) |
| ONNX | 支持ONNX Runtime的CPU/GPU环境 |
| RKNN | Rockchip RK3588开发板 |
快速开始
环境安装
pip install -r requirements.txt
Pytorch 推理
from transformers import AutoTokenizer, AutoModelForSequenceClassification
tokenizer = AutoTokenizer.from_pretrained("dean2023/cn-roberta-wwm-ext-car")
model = AutoModelForSequenceClassification.from_pretrained("dean2023/cn-roberta-wwm-ext-car")
inputs = tokenizer("发动机异响可能的原因有哪些？", return_tensors="pt")
outputs = model(**inputs)
RK3588 推理
from rknnlite.api import RKNNLite
rknn = RKNNLite()
rknn.load_rknn('./models/cn_roberta_v1.rknn')
rknn.init_runtime()
If you want to go further, please refer to the code.
Model info
Created: 2025-02-22
Updated: 2025-02-23
View on Hugging FaceFiles (8)
Related models
More from dean2023
No other models by this author.