TobyYang comments

Repositories
Issues
Comments

Results 2 comments of


                                            TobyYang

ollama can not run the custom model (finetune on llama3) on M1 max

❯ ollama create InsuranceGPT -f Modelfile transferring model data creating model layer creating template layer using already created layer sha256:9ff789dfa477f6c77a5e6a8e4be66df28d824c18cf84e24d1b14aac2d067e950 writing layer sha256:26fc224b78fce98e2c791de92ec7687171a7b46236f34faa64227cf9c2cfa4ac writing layer sha256:8854ff53cb2acccb83c87e946efc533a82b038d6d243caddd7ed0ba2e766ef84 writing manifest success ❯...

[Bug]: Running llama2-7b on H20, Floating point exception (core dumped) appears on float16

I met the same bug recently on H20. My torch version is: 2.3.1+cu121 You need to check your `conda list | grep cublas` first. Ensure there is only one `cublas`...