NLP & Transformers

Grape Models

A collection of T5 expert models hosted on Hugging Face focused on translating natural language into actionable terminal commands, routed by an intelligent semantic categorizer.

View Collection Back to Portfolio

Architecture & Base Model

The Grape Models ecosystem utilizes a Mixture of Experts (MoE) approach to maximize terminal command accuracy.

  • Base Foundation: All 4 expert models share the same foundation: Salesforce/codet5-small, allowing for lightweight execution.
  • Semantic Router: The minilm-l12-grape-route model acts as a traffic controller, evaluating the user's natural language input and routing it to the appropriate specialized expert.
  • Custom Fine-Tuning: Each expert was strictly fine-tuned on custom datasets using Google Colab instances for specialized tasks.
Explore the Router Model →
# NLP Pipeline Concept def process_input(text): # 1. Router categorizes intent category = router_model.predict(text) # 2. Select specific Grape Expert expert = load_expert(category) # 3. Generate terminal command command = expert.generate(text) return command

The Expert Models

Each model in the Grape family specializes in a specific domain of terminal operations.

Chardonnay BASH

The generalist expert. Converts natural language instructions into standard bash commands for file management, networking, and system operations.

Model Card →

Pinot SEARCH

The search and retrieval expert. Enforces a strict Zero-Modification Policy (ignores install/delete requests), focusing exclusively on information retrieval using tools like find, grep, and stat.

Model Card →

Malbec DOCKER

The containerization expert. Fine-tuned for Docker and docker-compose workflows. Trained on the custom public dataset malbec-nl2docker-es.

Model Card → Dataset →

Syrah NETWORK

The network expert. Specifically trained to handle network-specific tasks including port scanning, connectivity testing, DNS resolution, and firewall configuration. It understands complex context constraints.

Model Card →