Gen AI for Quant Fin Python Modeling 101 Hands-on using BERT

Python & Generative AI 101 for Freshmen

On this course we do Arms-on Gen AI for Quant Fin Python Modeling 101. We regularly repair our code utilizing ChatGPT which has similarities to Copilot on Python however since we use pocket book we use ChatGPT.

Python Generative AI for Modeling with ChatGpt & Copilot. GPT-Powered Chat Interface for querying rerunning tunning utilizing guide config for Hugging is confirmed. This Superb Tuning and connecting Chat characteristic to Logistic Regression backend will be prolonged with higher merchandise like Open AI. We question an already saved logistic regression mannequin with a GPT-powered chat interface to retrain and alter options and different modifications.

Course revolves round two tasks:

1. Superb-Tuning BERT with a Logistic Regression Layer
2. Deploying Fashions for Actual-Time Analytics: The way to use instruments like Flask/FastAPI to serve a text-based or data-generating mannequin.

Matters Launched

1. Intro to BERT vs GPT
2. Intro to Torch and Tensors
3. Intro to FastAPI App
4. In-memory Logistic Regression mannequin
5. Intro to  transformers like Coach, TrainingArguments, BertTokenizer, BertForSequenceClassification

Intro to Gen AI in Finance

1. Intro to BERT Fashions
2. Hugging face pre educated fashions, on-line account, native coaching
3. Intro to Superb tuning fashions, utilizing the sunshine DistilBERT, native coaching
4. Utilizing hugging face to map instructions for backend to question simulated outcomes, re-run simulation
5. Connecting Logistic regression to entrance finish of chat
6. Limitation of hugging face as of right this moment (Dec 2024
7. Arrange OpenAI API and hugging face

Future Work:

1. Characteristic Choice Mannequin Retraining on Actual Information
2. Intro to software of Gen AI in Information Analytics (synthesis anamoly and detection). Purposes of Generative AI in Analytics: Information synthesis, anomaly detection, and predictive modeling.
3. Utilizing open to get all spectrum of instruction to question Mannequin (we used 4-5 circumstances guide)

For Logsitic regression we perceive what all can we do. Outline the enter schema for retraining. Outline the coaching arguments with optimizations. We usse libraries like joblib or pickle to save lots of and cargo the logistic regression mannequin. Postman, cURL, or any shopper able to sending HTTP POST requests

Future work: Different Enhancements equivalent to including enter validation, characteristic monitoring, hyperparameter validation, and returning chances for predictions.