README.md

metadata

license: unknown
tags:
  - mistral
  - financial-sentiment
  - sentiment-analysis
  - finance
  - text-classification
  - lora
  - transformers
  - peft
model-index:
  - name: Mistral-7B Financial Sentiment Analyzer
    results:
      - task:
          type: text-classification
          name: Sentiment Analysis
        dataset:
          name: Financial Phrasebank
          type: financial_phrasebank
        metrics:
          - type: accuracy
            value: 1
            name: Accuracy

Mistral-7B Financial Sentiment Analyzer

Mistral-7B Financial Sentiment Analyzer is a fine-tuned version of mistralai/Mistral-7B-v0.1, designed specifically for financial sentiment analysis. It classifies financial texts into positive, negative, or neutral categories using a lightweight and efficient LoRA-based adaptation.

---

Model Details

• Model Name: diwakartiwari/mistral-7b-financial-sentiment
• Base Model: mistralai/Mistral-7B-v0.1
• Fine-tuning Method: LoRA (Low-Rank Adaptation)
• Trained Parameters: 41.9M (approximately 0.58% of 7.28B)
• Developer: Diwakar Tiwari
• Dataset: Financial Phrasebank (2,264 annotated samples)
• Task: Financial Sentiment Classification
• Accuracy: 100% on test cases

---

Use Cases

• Financial news sentiment analysis
• Investment research automation
• Risk assessment for financial institutions
• Market sentiment monitoring
• Integration into trading and research tools

---

Capabilities

• Detects sentiment polarity in financial news and statements
• Understands nuanced economic and corporate language
• Performs reliably on both short phrases and longer reports

---

Example Usage

from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel

# Load tokenizer and model
tokenizer = AutoTokenizer.from_pretrained("diwakartiwari/mistral-7b-financial-sentiment")
base_model = AutoModelForCausalLM.from_pretrained("mistralai/Mistral-7B-v0.1")
model = PeftModel.from_pretrained(base_model, "diwakartiwari/mistral-7b-financial-sentiment")

# Prepare prompt
prompt = "<s>[INST] Analyze the financial sentiment of this statement: Company reported record profits [/INST] "

# Generate response
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=20)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

# Expected Output: "The financial sentiment is positive."