prakhar146/financial-llm-cpu
prakhar146/financial-llm-cpu is a 1.5 billion parameter instruction-tuned causal language model, fine-tuned by prakhar146 on Qwen/Qwen2.5-1.5B-Instruct. It specializes in financial analysis, offering capabilities in technical analysis, trading decision-making, portfolio allocation, and macroeconomic analysis. This model is specifically optimized for CPU deployment and includes India-specific financial context, making it suitable for financial applications requiring local processing.
Loading preview...
Overview
prakhar146/financial-llm-cpu is a specialized financial large language model, built upon the Qwen/Qwen2.5-1.5B-Instruct architecture. It has been fine-tuned by prakhar146 using a substantial dataset of 74,998 financial samples, combining FinGPT, FinQA, and synthetic Chain-of-Thought data. The training utilized QLoRA and resulted in a low training loss of 0.1089, indicating effective learning.
Key Capabilities
- Technical Analysis: Provides step-by-step reasoning for market trends.
- Trading Decision Making: Assists in formulating trading strategies, outperforming GPT-4 in specific benchmarks (75.0% vs 72%).
- Portfolio Allocation & Risk Management: Supports strategic investment planning.
- Macroeconomic Analysis: Capable of analyzing broader economic indicators.
- India-Specific Context: Incorporates financial nuances relevant to the Indian market.
Performance Highlights
- Sentiment Analysis: Achieves 80.0% accuracy.
- Trading Decisions: Scores 75.0%.
- Reasoning Quality: Rated 70/100.
CPU Optimization
This model is designed for efficient deployment on CPU, making it accessible for environments without dedicated GPU resources. It can be loaded directly using AutoModelForCausalLM and AutoTokenizer from the Hugging Face transformers library with device_map="cpu" and torch_dtype=torch.float32.