Skip to content

MODEL_CARD.md

Latest commit

 

History

History
26 lines (21 loc) · 666 Bytes

MODEL_CARD.md

File metadata and controls

26 lines (21 loc) · 666 Bytes

Demo LoRA Adapter — Model Card

Overview

This adapter was trained using a tiny in‑memory corpus on an AMD RX 7700 XT using ROCm 7.2.x.
It is intended only as a demonstration of the training pipeline.

Base Model

  • microsoft/Phi-3-mini-4k-instruct

Training Details

  • 3 epochs
  • BF16 compute with 4-bit quantization (QLoRA)
  • LoRA (r=8, alpha=16, dropout=0.05)
  • Target modules: q_proj, v_proj
  • No external datasets
  • Training time: ~2 seconds (on demo corpus)

Intended Use

  • Demonstration
  • Testing ROCm compatibility
  • Educational purposes

Not Intended For

  • Production use
  • Safety‑critical applications
  • High‑quality text generation