61 lines
1.8 KiB
Markdown
61 lines
1.8 KiB
Markdown
---
|
|
base_model:
|
|
- Qwen/Qwen3-0.6B
|
|
license: mit
|
|
language:
|
|
- en
|
|
metrics:
|
|
- accuracy
|
|
pipeline_tag: text-classification
|
|
library_name: transformers
|
|
tags:
|
|
- finance
|
|
- sentiment-analysis
|
|
- lora
|
|
- qwen
|
|
- financial-news
|
|
---
|
|
# 🏦 Qwen3-0.6B - Financial Sentiment Classification (v2)
|
|
|
|
**Model name:** `YorkFr/financial-sentiment-qwen3-v2`
|
|
**Base model:** [Qwen/Qwen3-0.6B](https://huggingface.co/Qwen/Qwen3-0.6B)
|
|
**Fine-tuning type:** LoRA (merged full model)
|
|
**Task:** Financial news sentiment classification (`positive`, `neutral`, `negative`)
|
|
|
|
---
|
|
|
|
## 🧠 Overview
|
|
|
|
This model is a **fine-tuned version of Qwen3-0.6B**, specialized for **financial sentiment analysis**.
|
|
|
|
It classifies a short piece of financial or economic news into one of three categories:
|
|
|
|
- 🟢 `positive` — good market news, growth, profit, increase, etc.
|
|
- ⚪ `neutral` — balanced or uncertain tone.
|
|
- 🔴 `negative` — bad market news, loss, risk, decline, etc.
|
|
|
|
Training was performed using **LoRA (PEFT)** with a small balanced dataset of financial headlines and statements.
|
|
|
|
---
|
|
|
|
|
|
## 💡 Usage
|
|
|
|
You can load and use the model directly with 🤗 `transformers` — **no PEFT or special tokenizer required.**
|
|
|
|
```python
|
|
from transformers import AutoModelForCausalLM, AutoTokenizer
|
|
|
|
model = AutoModelForCausalLM.from_pretrained("YorkFr/financial-sentiment-qwen3-v2")
|
|
tokenizer = AutoTokenizer.from_pretrained("YorkFr/financial-sentiment-qwen3-v2")
|
|
|
|
text = (
|
|
"Instruction: Classify the sentiment of the following financial news sentence "
|
|
"as one of [positive, neutral, negative].\n"
|
|
"Sentence: Apple announces strong pre-orders.\n"
|
|
"Answer:"
|
|
)
|
|
|
|
inputs = tokenizer(text, return_tensors="pt")
|
|
outputs = model.generate(**inputs, max_new_tokens=5)
|
|
print(tokenizer.decode(outputs[0], skip_special_tokens=True)) |