malloc failed! on stories260 model

karpathy / llama2.c

Inference Llama 2 in one file of pure C

MIT License

17.15k stars 2.03k forks source link

Open vikramdattu opened 4 months ago

vikramdattu commented 4 months ago

I have cloned the tinyllamas repo and trying stories260 model. It fails with the malloc failed issue

./run tinyllamas/stories260K/stories260K.bin -z tinyllamas/stories260K/tok512.bin -t 0.0
malloc failed!

15M model however works out of the box. But not able to run any model from the GH repo!

Trying it on M1 Mac. Any ideas?

jameswdelancey commented 4 months ago

When you run tokenizer.py to export the tokenizer, the legacy version will work with run.c