license: mit
Gemma 2B IT - Customer Support Fine-tuned Model (QLoRA)
This model is a fine-tuned version of google/gemma-1.1-2b-it
using QLoRA on a custom instruction-tuning dataset designed for automating customer support tasks, including:
- ✉️ Complaint summarization
- 💬 Sentiment analysis
- 🧠 Topic modeling
- 📉 Churn prediction
- 🧾 Auto response drafting
🛠 Fine-tuning Details
- Technique: QLoRA (4-bit quantization using
bitsandbytes) - Dataset: 14k+ records combining Amazon review and Q&A data
- Data format: ChatML-style JSONL with
messages: [{role: ..., content: ...}] - Training platform: Google Colab (A100 GPU)
- Libraries: Hugging Face
transformers,peft,datasets
💡 Use Cases
This model is best suited for:
- Automating customer support replies using auto-response drafting
- Summarizing customer complaints to understand the needs better
- Classifying topics and customer sentiments
- Predicting churn based on interaction tone and topics
- Downloads last month
- 22