gemma4-e2b-cardholder-gguf

ai-colombia/gemma4-e2b-cardholder-gguf

gemma4-e2b-cardholder-gguf is a text generation model from ai-colombia. Use it when you need the model to write or continue text. It is set up for transformers. The card lists the license as gemma.

Fine-tuned google/gemma-4-E2B-it for on-device function calling in the AI Business Card Holder Android app.

Downloads · 30 days

8% of all-time downloads

All-time downloads

312

Public

Parameters

5.1B

13.7 GB on disk

Likes

Public

Hugging Face

Repo makeup

Click a slice to open those files.

.safetensors10.2 GB · 75%

At a glance

Task: Text Generation
Library: transformers
License: gemma
Model type: gemma4
Access: Public
Created: Apr 22, 2026
Updated: Apr 22, 2026
SHA: 525b0185

Try a prompt

Base models

google/gemma-4-E2B-it

Model card

From the Hugging Face model README

Gemma 4 E2B — Card Holder

Fine-tuned google/gemma-4-E2B-it for on-device function calling in the AI Business Card Holder Android app.

🔒 Private model — for use only inside the AI Business Card Holder app published by ai-colombia.

Intended use

Multilingual function calling for managing a contact / business-card database on a mobile phone:

find_contact(name, company?) — locate an existing contact
add_contact(name, phone?, email?, title?, company?) — create a new contact
update_contact(id, fields) — edit an existing contact
list_contacts(filter?) — list / search contacts
Free-form chat replies when no tool is appropriate

Languages

🇬🇧 English · 🇪🇸 Spanish · 🇫🇷 French · 🇵🇹 Portuguese · 🇩🇪 German · 🇳🇴 Norwegian

Out-of-scope use

❌ Any application other than the AI Business Card Holder app
❌ Redistribution, mirroring, or republication
❌ Further fine-tuning or model derivation
❌ Use in violation of the Gemma Terms of Use

Training


Base model	`google/gemma-4-E2B-it` (2.3 B effective / 5.1 B total)
Method	LoRA via Unsloth
Dataset	`ai-colombia/card-holder-finetune-data` (8 000 multi-turn examples, 6 languages) — private
Rank / α / dropout	16 / 16 / 0.05
Target modules	`q_proj`, `k_proj`, `v_proj`, `o_proj`, `gate_proj`, `up_proj`, `down_proj`
Effective batch size	16 (2 × 8 grad-accum)
Epochs / steps	2 / 1 000
Learning rate	5e-5, linear, 20 warmup steps
Optimizer	adamw_8bit
Precision	fp16 (Colab T4)
Sequence length	1024
Trainable params	31 M / 5.15 B (≈ 0.60 %)
Final training loss	0.7541
Train time	≈ 113 min on Colab T4
Date	April 2026

How to use

ℹ️ For mobile deployment, use the GGUF Q4_K_M quantization (see Files and versions once uploaded). The merged fp16 weights here are the source for that quantization.

llama.cpp / llama.rn / MediaPipe (Android)

<bos><start_of_turn>user
You are the Card Holder assistant. Use tools when needed.

Find John Smith from Microsoft.<end_of_turn>
<start_of_turn>model

Task: Text Generation
Library: transformers
Type: gemma4
License: gemma
Languages: en, es, fr, pt, de, no
Created: Apr 22, 2026
Updated: Apr 22, 2026