Create README.md

README.md

@@ -0,0 +1,33 @@
+---
+license: apache-2.0
+language:
+- en
+tags:
+- tiny-gpt
+- from-scratch
+- spiceechat
+pipeline_tag: text-generation
+---
+
+<p align="center">
+  <img src="https://readme-typing-svg.demolab.com?font=Space+Grotesk&weight=600&size=22&pause=800&color=00E7FF&center=true&vCenter=true&width=600&lines=SpiceeChat%2FBasemini;4+Million+Parameters;Trained+on+Vibes+and+ML+Jargon" />
+</p>
+
+**Basemini** is a tiny GPT-style transformer trained from scratch on synthetic ML sentences. It doesn't know what 2+2 is. It doesn't care. It will generate something that sounds like it belongs in a research paper abstract anyway.
+
+Born from an experiment to see how small a model could be and still call itself "AI". The answer: 4M parameters and a dream.
+
+**Load it like any HF model**
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer
+model = AutoModelForCausalLM.from_pretrained("SpiceeChat/Basemini", trust_remote_code=True)
+tokenizer = AutoTokenizer.from_pretrained("SpiceeChat/Basemini", trust_remote_code=True)
+```
+
+It runs on CPU. It runs on T4. It runs on a potato.
+
+You ask it something. It gives you a sentence with the word "transformer" in it. That's the deal.
+
+Made with sweat, tears, and `torch_fallback=True`. Pull requests welcome. Donations not needed. Moral support accepted.
+
+<img src="https://capsule-render.vercel.app/api?type=waving&height=80&section=footer&color=0:00E7FF,100:8B5CF6"/>