A 125.6M parameter causal language model trained from scratch on WikiText-103 and TinyStories.
No pretrained weights. Custom BPE tokenizer.
1import librarian
2model = librarian.load("130m")
3print(model.generate("The history of Rome began", temperature=0.7, top_k=40))
1pip install torch tokenizers
2python generate.py --prompt "The history of Rome began" --temperature 0.7 --top_k 40
1# Better
2model.generate("In the beginning , the Roman Empire was founded by")
3model.generate("Once upon a time in a land far away , there lived a")
4
5# Too short — likely to drift
6model.generate("rome")
Lower temperature (0.7) and top_k (40) give more coherent output than the defaults.
