NotebookMLX

English | 简体中文

meta-llama/NotebookLlama

I ported NotebookLlama and
implemented it with MLX 🔥

It uses mlx-community/Qwen2.5-1.5B-Instruct-bf16 for
pre-processing the
PDF, mlx-community/Qwen2.5-14B-Instruct-4bit for
creating
transcripts, mlx-community/Qwen2.5-7B-Instruct-4bit for
rewrites, and lucasnewman/f5-tts-mlx for Text-to-Speech ⚡

Citing the NotebookLlama outline.

Step 1: Pre-process PDF:
Use mlx-community/Qwen2.5-1.5B-Instruct-bf16 to
pre-process the PDF and save it in a .txt file.

Step 2: Transcript Writer:
Use mlx-community/Qwen2.5-14B-Instruct-4bit to write a
podcast transcript from the text.

Step 3: Dramatic Re-Writer: Use
the mlx-community/Qwen2.5-7B-Instruct-4bit model to
make the transcript more dramatic.

Step 4: Text-To-Speech Workflow:
Use lucasnewman/f5-tts-mlx to generate
a conversational podcast.

The podcast audio

https://github.com/user-attachments/assets/c7cf2d2f-766f-4026-8442-c584f6a32292

Star History