Downloads · 30 days
0
repairwise-gemma4-e2b-lora
abdullahfasih/repairwise-gemma4-e2b-lora
repairwise-gemma4-e2b-lora is a machine learning model from abdullahfasih. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as apache-2.0.
Fine-tuned from google/gemma-4-e2b using Unsloth.
Access
Public
Updated May 9, 2026
Repo size
134 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors101 MB · 76%
At a glance
- License
- apache-2.0
- Access
- Public
- Created
- May 9, 2026
- Updated
- May 9, 2026
- SHA
- 60b0aaa0
Model card
From the Hugging Face model README
repairwise-gemma4-e2b-lora
Fine-tuned from google/gemma-4-e2b using Unsloth.
What is this model?
RepairWise Gemma is a domain-adapted version of Gemma 4 E2B for phone repair support and SMS scam detection, fine-tuned on 241 multilingual examples across 6 languages.
Languages
Spanish · English · Catalan · Arabic · Romanian · Urdu
Categories
23 phone repair and safety categories including:
- SMS phishing / scam detection (HIGH risk)
- Battery swelling / fire risk (HIGH risk)
- Water damage emergency response (HIGH risk)
- Screen, charging, boot, wifi, app, storage issues (MEDIUM/LOW)
Fine-tuning details
- Base model: google/gemma-4-e2b (2B parameters)
- Method: LoRA (r=16, alpha=16) via Unsloth
- Dataset: 241 curated multilingual examples (RepairWise knowledge base)
- Training: 3 epochs, batch size 8, cosine LR schedule
- Hardware: Kaggle 2× T4 GPU
Benchmark results
| Metric | Score |
|---|---|
| Structure compliance | 90.0% |
| Urgency accuracy | 90.0% |
| Overall score | 90.0% |
Usage
from unsloth import FastModel
from unsloth.chat_templates import get_chat_template
model, tokenizer = FastModel.from_pretrained(
"abdullahfasih/repairwise-gemma4-e2b-lora",
max_seq_length=2048,
load_in_4bit=True,
)
tokenizer = get_chat_template(tokenizer, chat_template="gemma-4-thinking")
License
Apache 2.0
Part of
Gemma 4 Good Hackathon 2026 Track: Digital Equity & Inclusivity + Safety & Trust + Unsloth Special Prize
- License
- apache-2.0
- Created
- May 9, 2026
- Updated
- May 9, 2026