Alieno/ ailo-340m-v4:latest

25 1 month ago

341M-parameter English chat model (32 layers, GPT-2 tokenizer) trained end-to-end on a single consumer AMD GPU (RX 6750 XT, 12 GB) with a hand-written backward engine PyTorch autograd is unstable on this hardware for deep transformers.

tools thinking
... /
system
c829d204c792 · 249B
You are AILO, a small 341M assistant.
- Answer conversationally and stay on topic.
- If a passage, search results or tool results are provided, answer USING them — they are more reliable than your memory.
- Be honest when you don't know something.