diff --git a/docs/hub/_toctree.yml b/docs/hub/_toctree.yml index fbe31d443..d5b2f0405 100644 --- a/docs/hub/_toctree.yml +++ b/docs/hub/_toctree.yml @@ -333,9 +333,11 @@ - local: enterprise-hub-datasets title: Dataset viewer for Private datasets - local: enterprise-hub-resource-groups - title: Resource Groups (Advanced Access Control) + title: Resource Groups (Access Control) - local: advanced-compute-options title: Advanced Compute Options + - local: enterprise-hub-advanced-security + title: Advanced Security - local: enterprise-hub-tokens-management title: Tokens Management - local: enterprise-hub-analytics diff --git a/docs/hub/advanced-compute-options.md b/docs/hub/advanced-compute-options.md index c1d2e1ab0..4f42d9b04 100644 --- a/docs/hub/advanced-compute-options.md +++ b/docs/hub/advanced-compute-options.md @@ -6,6 +6,19 @@ Enterprise Hub organizations gain access to advanced compute options to accelera This feature is part of the Enterprise Hub. +## Host ZeroGPU Spaces in your organization + +ZeroGPU is a dynamic GPU allocation system that optimizes AI deployment on Hugging Face Spaces. By automatically allocating and releasing NVIDIA A100 GPUs (40GB VRAM) as needed, organizations can efficiently serve their AI applications without dedicated GPU instances. + +**Key benefits for organizations** + +- **Free GPU Access**: Access powerful NVIDIA A100 GPUs at no additional cost through dynamic allocation +- **Enhanced Resource Management**: Host up to 50 ZeroGPU Spaces for efficient team-wide AI deployment +- **Simplified Deployment**: Easy integration with PyTorch-based models, Gradio apps, and other Hugging Face libraries +- **Enterprise-Grade Infrastructure**: Access to high-performance NVIDIA A100 GPUs with 40GB VRAM per workload + +[Learn more about ZeroGPU →](https://huggingface.co/docs/hub/spaces-zerogpu) + ## Train on NVIDIA DGX Cloud Train on NVIDIA DGX Cloud offers a simple no-code training job creation experience powered by Hugging Face AutoTrain and Hugging Face Spaces. Instantly access NVIDIA GPUs and avoid the time-consuming work of writing, testing, and debugging training scripts for AI models. @@ -18,30 +31,30 @@ Read the [blogpost for Train on NVIDIA DGX Cloud](https://huggingface.co/blog/tr #### Transformers -| Architecture | -|---------------------------| -| Llama | -| Falcon | -| Mistral | -| Mixtral | -| T5 | -| gemma | +| Architecture | +| ------------ | +| Llama | +| Falcon | +| Mistral | +| Mixtral | +| T5 | +| gemma | #### Diffusers -| Architecture | -|---------------------------| -| Stable Diffusion | -| Stable Diffusion XL | +| Architecture | +| ------------------- | +| Stable Diffusion | +| Stable Diffusion XL | ### Pricing Usage of Train on NVIDIA DGX Cloud is billed by the minute of the GPU instances used during your training jobs. Usage fees accrue to your Enterprise Hub Organizations’ current monthly billing cycle, once a job is completed. You can check your current and past usage at any time within the billing settings of your Enterprise Hub Organization. -| NVIDIA GPU | GPU Memory | On-Demand Price/hr | -|---------------------------|---------------------------|---------------------------| -| NVIDIA L40S | 48GB |$2.75 | -| NVIDIA H100 | 80GB |$8.25 | +| NVIDIA GPU | GPU Memory | On-Demand Price/hr | +| ----------- | ---------- | ------------------ | +| NVIDIA L40S | 48GB | $2.75 | +| NVIDIA H100 | 80GB | $8.25 | ## NVIDIA NIM API (serverless) @@ -59,8 +72,8 @@ You can find all supported models in [this NVIDIA Collection](https://huggingfac Usage of NVIDIA NIM API (serverless) is billed based on the compute time spent per request. Usage fees accrue to your Enterprise Hub Organizations’ current monthly billing cycle, once a job is completed. You can check your current and past usage at any time within the billing settings of your Enterprise Hub Organization. -| NVIDIA GPU | GPU Memory | On-Demand Price/hr | -|---------------------------|---------------------------|---------------------------| -| NVIDIA H100 | 80GB |$8.25 | +| NVIDIA GPU | GPU Memory | On-Demand Price/hr | +| ----------- | ---------- | ------------------ | +| NVIDIA H100 | 80GB | $8.25 | -The total cost for a request will depend on the model size, the number of GPUs required, and the time taken to process the request. For each model, you can find which hardware configuration is used in the notes of [this NVIDIA Collection](https://huggingface.co/collections/nvidia/nim-66a3c6fcdcb5bbc6e975b508). \ No newline at end of file +The total cost for a request will depend on the model size, the number of GPUs required, and the time taken to process the request. For each model, you can find which hardware configuration is used in the notes of [this NVIDIA Collection](https://huggingface.co/collections/nvidia/nim-66a3c6fcdcb5bbc6e975b508). diff --git a/docs/hub/audit-logs.md b/docs/hub/audit-logs.md index 110e02d5f..0babbf71c 100644 --- a/docs/hub/audit-logs.md +++ b/docs/hub/audit-logs.md @@ -6,6 +6,76 @@ This feature is part of the Enterprise Hub. + + +Enterprise Hub organizations can improve their security with advanced security controls for both members and repositories. + +
+
+
@@ -38,16 +38,16 @@ Any repo (model or dataset) stored in a non-default location will display its Re
## Regulatory and legal compliance
-In regulated industries, companies may be required to store data in a specific region.
+Regulated industries often require data storage in specific regions.
-For companies in the EU, that means you can use the Hub to build ML in a GDPR compliant way: with datasets, models and inference endpoints all stored within EU data centers.
+For EU companies, you can use the Hub for ML development in a GDPR-compliant manner, with datasets, models and inference endpoints stored in EU data centers.
## Performance
-Storing your models or your datasets closer to your team and infrastructure also means significantly improved performance, for both uploads and downloads.
+Storing models and datasets closer to your team and infrastructure significantly improves performance for both uploads and downloads.
-This makes a big difference considering model weights and dataset files are usually very large.
+This impact is substantial given the typically large size of model weights and dataset files.

-As an example, if you are located in Europe and store your repositories in the EU region, you can expect to see ~4-5x faster upload and download speeds vs. if they were stored in the US.
+For example, European users storing repositories in the EU region can expect approximately 4-5x faster upload and download speeds compared to US storage.