| title: Ep. 7 — Tokenization & Embeddings | |
| emoji: 🔤 | |
| colorFrom: gray | |
| colorTo: purple | |
| sdk: static | |
| pinned: false | |
| license: mit | |
| # Ep. 7 — Tokenization & Embeddings | |
| How does a sentence become a list of vectors — and what does that space look like? | |
| ## Contents | |
| - **Notebook:** `tokenization_and_embeddings.ipynb` | |
| ## Sections | |
| | # | Topic | | |
| |---|-------| | |
| | 0 | Full pipeline — text → tokens → atom (live vLLM inference) | | |
| | 1 | BPE tokenization from scratch | | |
| | 2 | The embedding matrix — loading Qwen's actual weights | | |
| | 3 | Exploring embedding space (t-SNE, cosine similarity, nearest neighbors) | | |
| ## Hardware | |
| The notebook connects to the live vLLM server on atom (DGX Spark, Grace-Blackwell) for the inference section. Tokenizer and embeddings are loaded directly from Qwen/Qwen3.6-35B-A3B weights. | |
| ## Related | |
| - [Org hub](https://huggingface.co/EXD-AI) | |
| - [GitHub](https://github.com/Ramshreyas/EXD) | |