You can deploy this model and start earning money today!

ibokajordan/
RAG_llama2-7b-npc-dialogue

Name: RAG_llama2-7b-npc-dialogue
Availability: InStock

safetensors

Views

No views yet

RAG Llama 2 7B NPC Dialogue

This repository contains outputs from a RAG + QLoRA workflow using meta-llama/Llama-2-7b-chat-hf on a 20,000-sample training subset of the GameNPC dialogue dataset.

analysis_outputs/: training logs, train loss chart, error diagnostics, ablation results and sample question outputs.
final_adapter/: QLoRA adapter files, if the training cell was executed.
tokenizer/: tokenizer files saved from the active model session, if available.

Base model

meta-llama/Llama-2-7b-chat-hf

Experiment type

Retrieval-Augmented Generation (RAG)
QLoRA fine-tuning on 20,000 records
Error diagnostics
Ablation study
Sample question generation

Repository ID

ibokajordan/RAG_llama2-7b-npc-dialogue

Note

The base model is a gated Meta Llama model. Access must be approved on Hugging Face before running this notebook.

ibokajordan/RAG_llama2-7b-npc-dialogue