# @awomanindatascience on Instagram

- **Type:** Video
- **Original URL:** https://www.instagram.com/p/DTuO92NE0gC
- **Gondola URL:** https://gondola.cc/posts/61230117-awomanindatascience-instagram
- **Thumbnail:** https://img.gondola.cc/tr:w-,h-,fo-auto/postThumbnails/0ec1b8e9b3.jpg
- **Posted:** 2026-01-20T06:29:16.000+00:00
- **Account Owner:** Chinar Arora (@awomanindatascience) — https://gondola.cc/awomanindatascience

## Caption

Hello Folks! After a relaxing weekend it’s time to get into action for building the LLM ⚡️
For Day 17 of Building LLM from scratch, I implemented the full GPT architecture by stacking transformer blocks (12 layers for GPT-2 small).
Token + positional embeddings → dropout → multi-head attention + MLP blocks → LayerNorm → linear vocab projection.
Model outputs logits of shape (batch, seq_len, vocab_size) and supports autoregressive generation.
Untrained for now-training begins next.
#deeplearning #generativeai #womenwhocode #largelanguagemodels #womenintech

## Stats

- **Views:** 7,094
- **Likes:** 463
- **Shares:** 0
- **Comments:** 15

## Tags

womenintech, largelanguagemodels, deeplearning, womenwhocode, generativeai

---
Copyright (c) Gondola