Convert and PR models to Safetensors
ReWrite datasets with a text instruction
Organize and process datasets for AI models
Download datasets from a URL
Perform OSINT analysis, fetch URL titles, fine-tune models
Browse and view Hugging Face datasets from a collection
Search and find similar datasets
Upload files to a Hugging Face repository
Evaluate evaluators in Grounded Question Answering
Manage and orchestrate AI workflows and datasets
Create a domain-specific dataset project
Create a report in BoAmps format
Create a large, deduplicated dataset for LLM pre-training
Convert to Safetensors is a tool designed to convert AI models to the Safetensors format. Safetensors is a secure alternative to PyTorch's .pt files, offering enhanced security by only allowing the storage of floats and integers. This format is particularly useful for model sharing and collaboration, as it prevents potential attacks through malicious file content.
convert_to_safetensors --input /path/to/model.pt --output /path/to/model.safetensors
convert_to_safetensors --input /path/to/model.pt --output /path/to/model.safetensors --compress
Why should I use Safetensors over PyTorch's .pt files?
Safetensors provides enhanced security by restricting data types, reducing the risk of malicious content in shared models.
What formats are supported for conversion?
The tool supports conversion between PyTorch (.pt) and Safetensors (.safetensors) formats.
Can I convert a Safetensors model back to PyTorch?
Yes, the tool allows bidirectional conversion, enabling you to switch between formats as needed.