Customer Support LLM
LoRA fine-tune of Qwen/Qwen2-0.5B-Instruct trained with Hugging Face AutoTrain on rjac/e-commerce-customer-support-qa.
This prototype explores whether a small open-source language model can be adapted for e-commerce customer-support workflows relevant to SMEs.
Training Details
- Base model:
Qwen/Qwen2-0.5B-Instruct - Dataset:
rjac/e-commerce-customer-support-qa - Training column:
conversation - Method: LoRA / PEFT
- Platform: Hugging Face AutoTrain
- Examples: 1,000
- Epochs: 1
- Max sequence length: 1024
- Runtime: about 4 minutes on Nvidia T4
Intended Use
- customer-support response drafting
- e-commerce support workflow automation
- SME AI adoption prototype
- lightweight domain adaptation experiment
Example Use Case
A small e-commerce business could use a lightweight adapted model to draft first-pass support replies, help triage customer issues, and support customer-service workflows without relying only on closed frontier models.
Example Prompt
Customer: I ordered a phone case last week and the tracking page still says pending. Can you help?
Agent:
- Downloads last month
- -