Surasan/qwen2.5-7b-financial-thai-4bit-full

Qwen2.5-7B Financial Thai

This model is a fine-tuned version of Qwen2.5-7B-Instruct for Thai financial question answering.

Model Details

• Base Model: Qwen2.5-7B-Instruct
• Fine-tuning Method: LoRA (Low-Rank Adaptation)
• Training Data: Thai financial Q&A dataset
• Languages: Thai, English

Usage

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

# Load model and tokenizer
model_name = "Surasan/qwen2.5-7b-financial-thai-4bit-full"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype=torch.float16,
    device_map="auto"
)

# Generate response
def generate_answer(question):
    prompt = f'''<|im_start|>user
วิเคราะห์คำถามทางการเงินแล้วตอบพร้อมเหตุผลสั้นๆ

{question}<|im_end|>
<|im_start|>assistant
การคิด:'''
    
    inputs = tokenizer(prompt, return_tensors="pt")
    
    with torch.no_grad():
        outputs = model.generate(
            **inputs,
            max_new_tokens=256,
            temperature=0.1,
            do_sample=False,
            pad_token_id=tokenizer.eos_token_id,
        )
    
    response = tokenizer.decode(outputs[0][inputs['input_ids'].shape[1]:], skip_special_tokens=True)
    return response.strip()

# Example usage
question = "หุ้นของบริษัท A มีราคา 100 บาท PE ratio 15 เท่า บริษัท B มีราคา 80 บาท PE ratio 20 เท่า ควรเลือกลงทุนบริษัทใด?"
answer = generate_answer(question)
print(answer)

Training Details

• Training Steps: 300
• Learning Rate: 2e-4
• Batch Size: 8 (with gradient accumulation)
• LoRA Rank: 64
• LoRA Alpha: 128

Performance

Optimized for Thai financial question answering with reasoning capabilities.