Amsi-fin: Financial Vision-Language Model

Fine-tuned Qwen3-VL-4B for financial document understanding, chart analysis, and financial reasoning.

Quick Start

MLX (Apple Silicon)

from mlx_vlm import load, generate

# IMPORTANT: Use fix_mistral_regex=True
model, processor = load('AITRADER/Amsi-fin', fix_mistral_regex=True)

# Vision task
output = generate(
    model, processor, 
    image='chart.png',
    prompt='<|vision_start|><|image_pad|><|vision_end|>Analyze this chart.',
    max_tokens=500
)

# Text-only
output = generate(
    model, processor,
    prompt='Calculate debt-to-equity ratio if debt=120M, equity=80M.',
    max_tokens=200
)

Transformers (CUDA/CPU)

from transformers import AutoProcessor, AutoModelForVision2Seq
import torch

processor = AutoProcessor.from_pretrained('AITRADER/Amsi-fin', trust_remote_code=True)
model = AutoModelForVision2Seq.from_pretrained(
    'AITRADER/Amsi-fin',
    torch_dtype=torch.bfloat16,
    trust_remote_code=True
)

Capabilities

• Financial Document OCR
• Chart/Graph Understanding
• Financial Reasoning & Calculations
• Table Extraction

Training Data

• FinTrain (Salesforce)
• MultiFinBen-EnglishOCR
• ChartQA
• FinCoT