newest llama.cpp seems to crash now

#21
by WizardDave - opened

This is the newer model file you posted recently. This worked previously.

../llama.cpp/main --model ./models/ggml-vic13b-uncensored-q8_0.bin
main: build = 588 (ac7876a)
main: seed = 1684995027
llama.cpp: loading model from ./models/ggml-vic13b-uncensored-q8_0.bin
libc++abi: terminating due to uncaught exception of type std::runtime_error: unexpectedly reached end of file
zsh: abort ../llama.cpp/main --model ./models/ggml-vic13b-uncensored-q8_0.bin

Indeed, q4, 4_1, and 8 quantization formats have had breaking changes. Will update affected files asap

Thank you so much, I really like this model

FYI, 2d5db48 was the breaking change, last working commit is 6986c78

Having a ggml F16 version of the files would be nice, than we could also requantize when new changes are coming.

eachadea changed discussion status to closed
eachadea changed discussion status to open

Yeah, good idea. Will upload in a sec

Indeed, q4, 4_1, and 8 quantization formats have had breaking changes. Will update affected files asap

difference between q4, q4_1, q4_2 etc.
also: q4 = 4bits, q8 = 8bits is that right?

Correct. This is an obsolete model btw

Sign up or log in to comment