Auramind-90M - 90M Parameters

Ultra-lightweight for budget smartphones and edge devices

Specifications

• Parameters: 90M
• Base Model: google/gemma-2-270m
• Memory Usage: ~225MB RAM
• Quantization: INT8 optimized
• Inference Speed: 50-150ms on modern smartphones

Mobile Deployment

This variant is specifically optimized for:

• Target Devices: Budget smartphones and edge devices
• Memory Requirements: ~225MB RAM
• Performance: 50-150ms on modern smartphones

Usage

from transformers import AutoTokenizer, AutoModelForCausalLM

# Load this specific variant
tokenizer = AutoTokenizer.from_pretrained("zail-ai/auramind-90m")
model = AutoModelForCausalLM.from_pretrained(
    "zail-ai/auramind-90m",
    torch_dtype=torch.float16,
    device_map="auto",
    low_cpu_mem_usage=True
)

Refer to the main AuraMind repository for complete documentation.