emranyat/FinAI_adapters

FinAI Adapters

This repository contains fine-tuned LoRA adapters for financial AI tasks.

Adapters

finer_qwen

• Model: Qwen 3.8B with 8-bit quantization
• Task: Financial information extraction
• Rank: 8

xbrlNICL_qwen

• Model: Qwen 3.8B with 8-bit quantization
• Task: XBRL data processing with N-shot learning
• Rank: 8

Usage

Load the adapters using Hugging Face Transformers:

from transformers import AutoModelForCausalLM, AutoTokenizer

# Load base model
model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-3.8B")
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2.5-3.8B")

# Load adapter
model.load_adapter("emranyat/FinAI_adapters/finer_qwen")

Training Details

• Base Model: Qwen 3.8B
• Quantization: 8-bit
• LoRA Rank: 8
• Training Framework: Axolotl