The dataset is currently empty. Upload or create new data files. Then, you will be able to explore them in the Dataset Viewer.

Lingbot-VLA 大文件资产(配合 GitHub 代码仓)

本 Dataset 存放 未放进 GitHub 的大体积文件。代码与配置见:

目录结构

路径 说明 约大小
checkpoints/ 训练 checkpoint(含 global_step_1024/) ~32GB
pretrained/lingbot-vla-4b/ 预训练 VLA 权重 ~16GB
pretrained/Qwen2.5-VL-3B-Instruct/ 分词器 / VLM 权重 ~7GB
wheels/ flash-attn 预编译 wheel ~245MB
model_assets/ / runs/ 训练附属产物 较小

下载示例

pip install -U "huggingface_hub[cli]"
hf download coisini9293/lingbot --repo-type dataset --local-dir ./hf_lingbot_assets

按需下载单目录:

hf download coisini9293/lingbot checkpoints --repo-type dataset --local-dir ./checkpoints
hf download coisini9293/lingbot pretrained/lingbot-vla-4b --repo-type dataset --local-dir ./models/pretrained/lingbot-vla-4b

与本地路径对应

Dataset 路径 本机常见路径
checkpoints/ lingbot-vla/output/checkpoints/
pretrained/lingbot-vla-4b/ models/pretrained/lingbot-vla-4b/
pretrained/Qwen2.5-VL-3B-Instruct/ models/pretrained/Qwen2.5-VL-3B-Instruct/
wheels/*.whl 项目根目录 flash_attn-*.whl
Downloads last month
44