AiJoker/openchat_3.5-slp-jordan-belfort

Model Card for Jordan Belfort Q&A Model

This model is a fine-tuned version of a transformer-based language model trained using supervised fine-tuning (SFT) on a custom Q&A dataset derived from Jordan Belfort's book. The model is optimized to answer questions related to the book's content, including topics like sales, persuasion, mindset, and personal development strategies.

Model Details

• —Developed by: Jobix.ai
• —Finetuned from model: `openchat 3.5*
• —Language(s): English
• —Model type: Q&A / Instruction-following
• —License: apache-2.0 (or your chosen license)

Model Sources

• —Training Data: Custom Q&A dataset built from the full content of Jordan Belfort’s book.
• —Method: Supervised fine-tuning (TRL + SFT)

Uses

Direct Use

• —Ask specific questions about concepts, strategies, and advice in Jordan Belfort's book.
• —Get summaries of chapters, sales techniques, or mindset frameworks presented in the book.
• —Useful for salespeople, coaches, or individuals studying persuasion and personal development.

Out-of-Scope Use

• —Not trained for general-purpose Q&A outside the context of the book.
• —Not suitable for legal, financial, or medical advice.

Training Details

Training Procedure

• —Trainer: trl.SFTTrainer
• —Precision: bfloat16
• —Epochs: 7
• —Optimizer: AdamW
• —LR Scheduler: Cosine with warmup
• —Loss: CrossEntropyLoss on prompt-response pairs

Dataset

• —Approx. ~2,000 curated Q&A pairs covering all chapters and sections of the book.
• —Balanced across concepts like tonality, straight-line persuasion, mindset, sales process, and personal stories.

Evaluation

• —Manual evaluation on question coverage and accuracy.
• —Model shows strong performance in recalling specific ideas and quoting relevant sections.

Example Usage

from transformers import pipeline

qa = pipeline("text-generation", model="your-username/jordan-belfort-qa")

prompt = "What is the straight-line sales method according to Jordan Belfort?"
response = qa(prompt, max_new_tokens=200, do_sample=False)
print(response[0]["generated_text"])