"error loading model: missing tok_embeddings.weight" with llama.cpp
#5
by gardner - opened
Hi there,
thanks for publishing all these models. I just tried this one out with llama.cpp. I pulled from the repo and rebuilt with CUBLAS enabled and run with:
./build/bin/main -m /home/user/src/llama.cpp/models/WizardCoder-15B/WizardCoder-15B-1.0.ggmlv3.q4_1.bin -p "Please tell me a story about a robot."
I tried the q4_0.bin file as well with the same result.
$ git log -1
commit 049aa16b8c5c6d086246e4e6b9feb18de4fbd663 (HEAD -> master, origin/master, origin/HEAD)
Author: Georgi Gerganov <[email protected]>
Date: Tue Jun 20 19:05:54 2023 +0300
readme : add link to p1
They're not compatible with llama.cpp - please see the README for a list of software it will work with.
gardner changed discussion status to closed