FINRA-LM v2 1.5B — Financial Reasoning & Analysis Language Model
A domain-restricted financial AI fine-tuned from Qwen2.5-1.5B-Instruct using QLoRA. FINRA-LM v2 covers equities, investment banking, fixed income, derivatives, macro, sentiment analysis, and personal finance — and actively refuses off-topic queries.
What it can do
- Stock analysis: P/E, EV/EBITDA, DCF, technical analysis
- Investment banking: M&A, LBO, IPO, comparable company analysis
- Financial sentiment: positive / neutral / negative on news
- Earnings report analysis (numerical + textual)
- Macro + fixed income: rates, bonds, yield curves
- Derivatives: options, futures, Greeks
- Personal finance and regulatory compliance
What it will NOT do
Off-topic questions are gracefully rejected.
Training Details
| Parameter | Value |
|---|---|
| Base model | Qwen2.5-1.5B-Instruct |
| Method | QLoRA (r=16, alpha=16) via Unsloth |
| Datasets | fingpt-fiqa_qa, finance-alpaca, fingpt-sentiment-train, fingpt-convfinqa |
| Hardware | Google Colab T4 GPU |
GGUF for LM Studio
Download from: tanmaybhardwaj19/finra-lm-v2-gguf
Load the .gguf file directly in LM Studio for local inference.
Usage
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
model = AutoModelForCausalLM.from_pretrained('tanmaybhardwaj19/finra-lm-v2-1.5b')
tokenizer = AutoTokenizer.from_pretrained('tanmaybhardwaj19/finra-lm-v2-1.5b')
messages = [{"role": "user", "content": "Analyze Apple (AAPL) stock."}]
inputs = tokenizer.apply_chat_template(messages, return_tensors="pt", add_generation_prompt=True)
outputs = model.generate(inputs, max_new_tokens=512, temperature=0.7)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))