Dataset Viewer

The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.

AutoML Benchmarks Mirror

本仓库是几个常用 NAS 基准(NAS-Bench-101 / 201 / ASR)数据文件的镜像,目的是给 marsggbo/AutoMLDemos(《动手学 AutoML》配套代码)提供一份 稳定可用的备份,防止上游源失效。

创建动机

  • github.com/SamsungLabs/nb-asr 的 NAS-Bench-ASR 仓库已经被官方删除(截至 2026 年)。
  • NAS-Bench-201 的 nb201.pth 走 Google Drive,经常遇到配额或网络问题。
  • NAS-Bench-101 的 tfrecord 文件托管在 storage.googleapis.com/nasbench/,目前可用但未来不可保证。

📦 包含的数据

nasbench101/

文件 大小 来源
nasbench_only108.tfrecord ~520 MB https://storage.googleapis.com/nasbench/nasbench_only108.tfrecord
nasbench_full.tfrecord ~2 GB https://storage.googleapis.com/nasbench/nasbench_full.tfrecord

nasbench201/

文件 大小 来源
nb201.pth ~4.5 GB Google Drive id 16Y0UwGisiouVRxW-W5hEtbxmcHw_0hF_

nasbenchasr/ ⚠️ 数据缺失

原始仓库 github.com/SamsungLabs/nb-asr 已被删除,目前找不到完整镜像。如果你手头有以下文件, 欢迎 PR 到本仓库补全:

nb-asr-bench-gtx-1080ti-fp32.pickle
nb-asr-bench-jetson-nano-fp32.pickle
nb-asr-e10-1234.pickle
nb-asr-e40-1234.pickle
nb-asr-e40-1235.pickle
nb-asr-e5-1234.pickle
nb-asr-info.pickle

🚀 用法

Python

from huggingface_hub import hf_hub_download

path = hf_hub_download(
    repo_id="marsggbo/automl-benchmarks",
    filename="nasbench101/nasbench_only108.tfrecord",
    repo_type="dataset",
)

国内加速

export HF_ENDPOINT=https://hf-mirror.com

或在 Python 里 os.environ["HF_ENDPOINT"] = "https://hf-mirror.com"

直链

https://huggingface.co/datasets/marsggbo/automl-benchmarks/resolve/main/<path>

License & Attribution

本仓库只是上游数据的搬运,所有数据所有权和 License 归原作者。

  • NAS-Bench-101 © Google Research, Apache 2.0
  • NAS-Bench-201 © Xuanyi Dong, MIT
  • NAS-Bench-ASR © Samsung Labs, Apache 2.0

如果有任何版权问题请联系 marsggbo 删除。

Downloads last month
58