Skip to main content

AI Resource (Asset Hub)

AI Resource is your ultimate shortcut from raw idea to production-ready AI. We've eliminated the most frustrating bottlenecks in machine learning—waiting for massive weights to download, wrestling with data migration, and configuring broken environments.

Alaya NeW Cloud has pre-cached the world's best open-source foundation models, massive datasets, and optimized Docker images directly onto our high-speed network. With zero data migration required, you can seamlessly attach these assets to your sheer Compute and jump straight into fine-tuning, inference, or massive distributed training like a pro.

🚀 Key Features​

1️⃣ Model Repository (Skip the Download)​

Stop wasting hours pulling weights from the internet. The Model Repository is pre-loaded with top-tier, mainstream open-source models (e.g., Qwen2.5, Whisper-large, DINOv2, DistilBERT).

  • Out-of-the-box Injection: Models are instantly accessible inside your Cloud Container Instance (CCI) at the absolute path: /root/public/models. No downloads, no waiting.
  • Deep Categorization: Easily filter models by Tasks (Text-to-Image, Sentence Similarity, Object Detection, Speech Recognition), Libraries (PyTorch, Transformers.js, JAX, MLX), Languages, and Licenses.

2️⃣ Datasets (Zero-Migration Data Hub)​

Train your models without the bandwidth headache. We've pre-staged a massive library of industry-standard datasets (e.g., GSM8K, HumanEval, LibriSpeech) so you can seamlessly bind them to your compute instances.

  • Direct Path Access: Read massive datasets instantly inside your CCI at /root/public/datasets.
  • Multi-Modal Support: Filter datasets exactly how you need them—by Modalities (Text, Audio, Video, 3D, Geospatial), Formats (JSON, CSV, Parquet, WebDataset), or targeted ML Tasks.

3️⃣ Image Registry (Reproducible Runtimes)​

Your code needs a bulletproof environment to run safely and efficiently. The Image Registry manages all your Docker containers for rapid, scalable deployment.

  • Public Images: Tap into our pre-built, heavily optimized environments. These images come pre-installed with the necessary frameworks (PyTorch, DeepSpeed) to guarantee a secure, stable, and high-performance base for Foundation Model training.
  • Private Images: Snapshot your perfect workspace and save it to your Private Registry. Manage your team's custom Docker images, track storage Limit and Usage metrics (e.g., 8.41 GB used / 50.00 GB limit), and manage granular access via the Authorized Users list.