Jarvis-Astra (v2.0.0)

Architected & Trained by Yug Sisodiya using the J.A.R.V.I.S. Neural Forge studio.

🌟 Model Overview

• Model Name: Jarvis-Astra
• Release Version: v2.0.0
• Architecture: Y-AI Transformer
• Parameter Capacity: 0.52M Weights
• Context Window: 32,768 tokens
• Trained Weights Checkpoint: pytorch_model.bin (2.02 MB)

🚀 PyTorch Quickstart (Direct Weights Inference)

Load and run the trained weights in Python using PyTorch:

import torch
from model import YAIJarvisModel, YAIConfig

# 1. Initialize configuration
config = YAIConfig(
    d_model=64,
    num_layers=2,
    n_heads=4,
    vocab_size=128
)

# 2. Instantiate architecture & load checkpointed weights
model = YAIJarvisModel(config)
state_dict = torch.load("pytorch_model.bin", map_location="cpu")
model.load_state_dict(state_dict)
model.eval()

print("✓ Jarvis-Astra neural weights loaded successfully!")

🦙 Run with Ollama

ollama run hf.co/yugsisodiya/jarvis-jarvis-astra

🧠 System Architecture Highlights

• Checkpointed neural weights verified in local studio runtime.
• Low-latency interactive inference, multi-domain reasoning, and executive problem solving.
• Fully integrated with the J.A.R.V.I.S. unified cognitive assistant architecture on port 4700.