Postnet Qwen05b With Gates is an int8 ONNX language model designed for federated learning scenarios where browser tabs act as worker nodes and coordination is managed by a Cloudflare Durable Object. 5B-Instruct and has been exported for ORT-web compatibility, then quantized to int8 for efficient browser-based inference. It is distributed as a single-file ONNX model, approximately 906 MB in size, which streams directly from the Hugging Face Hub to client browsers.
The tool enables a federated learning approach in which each browser tab runs model inference using onnxruntime-web and participates in federated-SPSA training of a sparse NTK-Mirror gate controller. This controller features 5000 signed log-gates, and training occurs collaboratively among all connected users who open the provided demo URL. No installation or repository cloning is required, as the model is delivered directly to the browser for immediate participation in the distributed training process.
0 license. Its architecture and delivery method make it suitable for researchers and developers interested in browser-based federated learning and distributed model training using web technologies.
Postnet Qwen05b With Gates sits in PulseGate's Other AI category. It focuses on enabling federated learning and efficient browser-based inference with a quantized open-source language model. It is built as an open-source project for AI researchers and developers. Postnet Qwen05b With Gates is open source under the MIT license. Postnet Qwen05b With Gates is available on the web and the command line.
abgunaydin builds and maintains Postnet Qwen05b With Gates, and the product first shipped in 2026. Development happens publicly on GitHub with 83 commits in the last 90 days. Key capabilities include ONNX model, federated learning, and browser inference.
Latest indexed changes and source events
abgunaydin/postnet-qwen05b-with-gates verified by the PulseGate indexer
Other apps tracked under the same category.
