metadata
base_model: google/gemma-3-270m-it
tags:
- text-generation
- social-media-hashtags
- gemma
- Web4
- MLM
- Linkspreed
- Hashtags
license: apache-2.0
🤖 Web4/LS-W4-270M-Micro-Hashtags Model Card
💻 Model Details
- Model Name: Web4/LS-W4-270M-Micro-Hashtags
- Model Type: Micro Language Model (MLM)
- Model Size: 270 Million parameters (270M)
- Developer: Web4
- Intended Use: On-device, offline, and browser-based hashtag generation from user text input
- License: MIT
📄 Model Description
The LS-W4-270M-Micro-Hashtags is a compact, high-efficiency 270M parameter Micro Language Model (MLM) developed by Web4.
It is designed specifically for environments with limited computational resources, such as mobile devices, desktop apps, and browsers — with no server required.
Its primary function is to generate relevant, trending, and semantically meaningful hashtags from input text in real time.
It runs fully offline, supports WebAssembly/WebGPU, and enables privacy-preserving, decentralized AI applications.
📊 Training Data and Procedure
- Training Dataset Size: Over 1 million diverse datasets
- Data Sources: Social media posts, online articles, and real-world text snippets tagged with hashtags
- Objective: Masked Language Modeling (MLM) + fine-tuning for hashtag prediction
- Performance Optimizations: Designed for fast inference, low latency, and minimal memory footprint
⚙️ Technical Specifications
| Specification | Value |
|---|---|
| Parameters | 270M |
| Model Format | ONNX, Web4 Custom Binary, WebAssembly |
| Operating Environment | On-device (mobile/desktop), Browser (WASM/WebGPU) |
| Server Requirement | None (fully offline) |
| Inference Speed | Real-time on consumer hardware |
| Primary Function | Hashtag generation from freeform text |